<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.4 20241031//EN" "JATS-journalpublishing1-4.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="research-article" dtd-version="1.4" xml:lang="en">
  <front>
    <journal-meta>
      <journal-id journal-id-type="publisher-id">am</journal-id>
      <journal-title-group>
        <journal-title>Applied Mathematics</journal-title>
      </journal-title-group>
      <issn pub-type="epub">2152-7393</issn>
      <issn pub-type="ppub">2152-7385</issn>
      <publisher>
        <publisher-name>Scientific Research Publishing</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="doi">10.4236/am.2026.1711040</article-id>
      <article-id pub-id-type="publisher-id">am-155388</article-id>
      <article-categories>
        <subj-group>
          <subject>Article</subject>
        </subj-group>
        <subj-group>
          <subject>Physics</subject>
          <subject>Mathematics</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Reconstruction of Missing Well-Log DTC Curves Using Low-Rank Matrix Recovery</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Li</surname>
            <given-names>Haitao</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author" corresp="yes">
          <name name-style="western">
            <surname>Fang</surname>
            <given-names>Yizhu</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Chen</surname>
            <given-names>Yanru</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Zou</surname>
            <given-names>Yu</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <name name-style="western">
            <surname>Zheng</surname>
            <given-names>Shu</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
      </contrib-group>
      <aff id="aff1"><label>1</label> Planning Department, PetroChina Southwest Oil and Gas Field Company, Chengdu, China </aff>
      <author-notes>
        <fn fn-type="conflict" id="fn-conflict">
          <p>The authors declare no conflicts of interest regarding the publication of this paper.</p>
        </fn>
      </author-notes>
      <pub-date pub-type="epub">
        <day>13</day>
        <month>11</month>
        <year>2026</year>
      </pub-date>
      <pub-date pub-type="collection">
        <month>11</month>
        <year>2026</year>
      </pub-date>
      <volume>17</volume>
      <issue>11</issue>
      <fpage>720</fpage>
      <lpage>743</lpage>
      <history>
        <date date-type="received">
          <day>11</day>
          <month>10</month>
          <year>2026</year>
        </date>
        <date date-type="accepted">
          <day>15</day>
          <month>11</month>
          <year>2026</year>
        </date>
        <date date-type="published">
          <day>18</day>
          <month>11</month>
          <year>2026</year>
        </date>
      </history>
      <permissions>
        <copyright-statement>© 2026 by the authors and Scientific Research Publishing Inc.</copyright-statement>
        <copyright-year>2026</copyright-year>
        <license license-type="open-access">
          <license-p> This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ( <ext-link ext-link-type="uri" xlink:href="https://creativecommons.org/licenses/by/4.0/">https://creativecommons.org/licenses/by/4.0/</ext-link> ). </license-p>
        </license>
      </permissions>
      <self-uri content-type="doi" xlink:href="https://doi.org/10.4236/am.2026.1711040">https://doi.org/10.4236/am.2026.1711040</self-uri>
      <abstract>
        <p>To address partial missing values in compressional sonic travel time (DTC) curves, this study develops a missing-log reconstruction method based on low-rank matrix recovery. First, multiple well-log variables, including gamma ray, bulk density, neutron porosity, resistivity, total porosity, water saturation, and DTC, are integrated into a standardized multi-log observation matrix. Considering the correlations among different well-log variables and the vertical continuity of logging responses along depth, a depth-wise smoothness constraint is introduced to improve the geological consistency of the reconstructed DTC curves. Artificial masking experiments are conducted on complete well-log datasets to evaluate the reconstruction performance of the proposed method. Isolated missing points and continuous missing intervals are designed separately to simulate different missing-data patterns in practical logging applications. The proposed method is compared with baseline methods, including linear interpolation and pure low-rank recovery. The experimental results show that the proposed method achieves better reconstruction accuracy for missing DTC curves, while effectively preserving the overall variation trend and depth continuity of the curves. These results indicate that combining low-rank matrix recovery with a depth-continuity prior provides an effective and geologically reasonable solution for missing well-log reconstruction.</p>
      </abstract>
      <kwd-group kwd-group-type="author-generated" xml:lang="en">
        <kwd>Well-Log Data</kwd>
        <kwd>Missing Data Reconstruction</kwd>
        <kwd>Low-Rank Matrix Recovery</kwd>
        <kwd>Compressional Sonic Travel Time</kwd>
        <kwd>DTC</kwd>
        <kwd>Depth Continuity</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec1">
      <title>1. Introduction</title>
      <p>In recent years, the reconstruction of missing well-log data has been investigated using a variety of data-driven approaches. Traditional interpolation-based methods are still widely used in practical applications because of their simple implementation and low computational cost [<xref ref-type="bibr" rid="B1">1</xref>]. However, these methods mainly rely on local continuity along the depth direction and reconstruct missing values only from neighboring samples of the same log. As a result, they are unable to fully exploit the correlations among different logging curves. Their performance may deteriorate when the missing interval is long, when abrupt lithological changes occur, or when the log response is controlled by nonlinear relationships among multiple geological and petrophysical factors. In such cases, interpolation may produce overly smooth curves and fail to recover the true local variation of the missing log.</p>
      <p>Recent studies further show that missing well-log imputation remains an active research topic. For example, benchmark comparisons of machine-learning methods indicate that reconstruction performance may vary with the selected model, logging variables, and missing-data characteristics [<xref ref-type="bibr" rid="B2">2</xref>]. In addition, collaborative-filtering-based imputation has been introduced to exploit similarity information among logging samples, providing another data-driven perspective for missing-value reconstruction [<xref ref-type="bibr" rid="B3">3</xref>]. These studies suggest that missing well-log reconstruction is not only a simple interpolation problem, but also a data-driven estimation problem that depends on both inter-log relationships and the structure of missing patterns.</p>
      <p>To overcome these limitations, machine learning methods have been introduced to predict missing logging curves by learning nonlinear relationships from available well-log data. Early studies used artificial neural networks and related machine-learning techniques to estimate petrophysical properties or generate synthetic well logs from existing logging measurements [<xref ref-type="bibr" rid="B4">4</xref>]-[<xref ref-type="bibr" rid="B7">7</xref>]. More recently, deep-learning-based methods have also been applied to missing well-log prediction and have shown promising performance in real reservoir cases [<xref ref-type="bibr" rid="B8">8</xref>] [<xref ref-type="bibr" rid="B9">9</xref>]. Compared with interpolation-based methods, machine learning approaches are better able to capture nonlinear mapping relationships among multiple logging curves. However, these methods usually treat samples at different depths as relatively independent observations, and the continuity constraint along the depth direction within the same well is often insufficiently considered. Therefore, in continuous missing intervals or in formations with complex geological variations, the predicted results may exhibit discontinuous fluctuations along the depth direction, which can affect the geological plausibility and morphological consistency of the reconstructed curve. Deep learning models, including convolutional neural networks, recurrent neural networks, spatio-temporal neural networks, self-attention-based architectures, and graph neural networks, have also been applied to well-log prediction and reconstruction tasks [<xref ref-type="bibr" rid="B10">10</xref>]-[<xref ref-type="bibr" rid="B13">13</xref>]. Although these methods have strong representation capabilities and can model sequential dependencies to some extent, their performance usually depends on large amounts of training data and careful model tuning.</p>
      <p>To tackle these drawbacks of existing methods, this study puts forward a missing well-log reconstruction method that combines low-rank matrix recovery with depth-wise smoothness correction. The proposed method integrates multiple well-log curves into a unified observation matrix and exploits the inherent low-rank structure of multi-log data to estimate the missing parts of the target log. Besides, given that well-log responses in the same well usually change continuously as depth increases, we also add adepth-wise smoothness correction mechanism to make the reconstruction results more vertically consistent and geologically reasonable [<xref ref-type="bibr" rid="B14">14</xref>]. In short, the proposed method combines the correlation between different well-log curves and the vertical continuity within a single well, thereby improving the reliability and accuracy of missing well-log reconstruction.</p>
      <p>In this study, the DTC curves are selected as the main reconstruction target, while GR (Gamma Ray), RHOB (Bulk Density), NPHI (Neutron Porosity), RT (Resistivity), PHIT (Total Porosity), and water saturation (SW) are used as auxiliary logging variables. Artificial masking experiments are conducted on complete well-log datasets to evaluate the reconstruction performance [<xref ref-type="bibr" rid="B15">15</xref>]. The proposed method is assessed using mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), coefficient of determination (<inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> ), and correlation coefficient (Corr). In addition, comparative experiments against linear interpolation and pure low-rank recovery without smoothness correction are carried out to examine both the overall effectiveness of the low-rank framework and the contribution of the proposed depth smoothness correction. The results demonstrate that the proposed framework can achieve accurate, stable, and geologically plausible reconstruction for incomplete well-log data. </p>
    </sec>
    <sec id="sec2">
      <title>2. Preliminaries</title>
      <sec id="sec2dot1">
        <title>2.1. Multi-Log Data Representation</title>
        <p>In well-log interpretation, measurements obtained from multiple logging curves are typically integrated into a unified depth-by-variable matrix to provide support for reconstruction analysis. It is assumed that the dataset comprises <inline-formula><mml:math><mml:mi> n </mml:mi></mml:math></inline-formula> depth samples and <inline-formula><mml:math><mml:mi> p </mml:mi></mml:math></inline-formula> logging variables, where the selected variables include GR, RHOB, NPHI, DTC, RT, PHIT, and SW. These measurements are organized into an observation matrix <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> X </mml:mi></mml:mstyle><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> p </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> , in which each row corresponds to a depth sample and each column corresponds to a single logging curve [<xref ref-type="bibr" rid="B16">16</xref>]. The primary objective of this study is to reconstruct the missing values in the DTC log with the assistance of the auxiliary variables.</p>
        <p>Because different logging variables have different physical units and numerical ranges, the data matrix is preprocessed before reconstruction. Specifically, the selected variables are converted to numeric format, the RT variable is logarithmically transformed, and all variables are standardized. After preprocessing, the standardized matrix is denoted by <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> . To describe missing-data locations, a binary observation mask matrix <inline-formula><mml:math><mml:mi> Ω </mml:mi></mml:math></inline-formula> is introduced, where <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 1 </mml:mn></mml:mrow></mml:math></inline-formula> indicates that the entry is observed and <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 0 </mml:mn></mml:mrow></mml:math></inline-formula> indicates that it is missing.</p>
      </sec>
      <sec id="sec2dot2">
        <title>2.2. Low-Rank Modeling Background</title>
        <p>Based on the multi-log matrix representation introduced in Section 2.1, low-rank modeling provides a natural way to characterize the shared structure among different logging variables. In subsurface formations, logging responses are controlled by common geological and petrophysical factors, such as lithology, porosity, fluid content, and compaction. As a result, different logs are often strongly correlated rather than independent. This inter-variable dependency implies that the standardized multi-log matrix <inline-formula><mml:math><mml:mi> Z </mml:mi></mml:math></inline-formula> contains redundant information and can therefore be approximated by a low-rank structure.</p>
        <p>Formally, the singular value decomposition (SVD) of <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> can be written as </p>
        <disp-formula id="FD1">
          <mml:math>
            <mml:mrow>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>Z</mml:mi>
              </mml:mstyle>
              <mml:mo>=</mml:mo>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>U</mml:mi>
              </mml:mstyle>
              <mml:mi>Σ</mml:mi>
              <mml:msup>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>V</mml:mi>
                </mml:mstyle>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> U </mml:mi></mml:mstyle><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> n </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> V </mml:mi></mml:mstyle><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> p </mml:mi><mml:mo> × </mml:mo><mml:mi> p </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> are orthogonal matrices, and <inline-formula><mml:math><mml:mrow><mml:mi> Σ </mml:mi><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> p </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> is a diagonal matrix whose diagonal entries are the singular values of <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> arranged in non-increasing order.</p>
        <p>If the singular values decay rapidly, <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> can be well approximated by a rank-<inline-formula><mml:math><mml:mi> r </mml:mi></mml:math></inline-formula> truncated SVD:</p>
        <disp-formula id="FD2">
          <mml:math>
            <mml:mrow>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>Z</mml:mi>
              </mml:mstyle>
              <mml:mo>≈</mml:mo>
              <mml:msub>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>Z</mml:mi>
                </mml:mstyle>
                <mml:mi>r</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:msub>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>U</mml:mi>
                </mml:mstyle>
                <mml:mi>r</mml:mi>
              </mml:msub>
              <mml:msub>
                <mml:mi>Σ</mml:mi>
                <mml:mi>r</mml:mi>
              </mml:msub>
              <mml:msubsup>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>V</mml:mi>
                </mml:mstyle>
                <mml:mi>r</mml:mi>
                <mml:mo>⊤</mml:mo>
              </mml:msubsup>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:mi> r </mml:mi><mml:mo> ≪ </mml:mo><mml:mtext> min </mml:mtext><mml:mrow><mml:mo> ( </mml:mo><mml:mrow><mml:mi> n </mml:mi><mml:mo> , </mml:mo><mml:mi> p </mml:mi></mml:mrow><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> , <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> U </mml:mi></mml:mstyle><mml:mi> r </mml:mi></mml:msub><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> , <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Σ </mml:mi><mml:mi> r </mml:mi></mml:msub><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> r </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> , and <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> V </mml:mi></mml:mstyle><mml:mi> r </mml:mi></mml:msub><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> p </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> . According to the Eckart-Young theorem, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle><mml:mi> r </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the best rank-<inline-formula><mml:math><mml:mi> r </mml:mi></mml:math></inline-formula> approximation to <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> under the Frobenius norm.</p>
        <p>For the purpose of subsequent optimization, the truncated SVD can be equivalently expressed in a bilinear factorization form:</p>
        <disp-formula id="FD3">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>Z</mml:mi>
                <mml:mi>r</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mi>A</mml:mi>
              <mml:msup>
                <mml:mi>B</mml:mi>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where</p>
        <disp-formula id="FD4">
          <mml:math>
            <mml:mtable columnalign="left">
              <mml:mtr>
                <mml:mtd>
                  <mml:mi>A</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:msub>
                    <mml:mi>U</mml:mi>
                    <mml:mi>r</mml:mi>
                  </mml:msub>
                  <mml:msubsup>
                    <mml:mi>Σ</mml:mi>
                    <mml:mi>r</mml:mi>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mn>1</mml:mn>
                        <mml:mo>/</mml:mo>
                        <mml:mn>2</mml:mn>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msubsup>
                  <mml:mo>,</mml:mo>
                </mml:mtd>
              </mml:mtr>
              <mml:mtr>
                <mml:mtd>
                  <mml:mi>B</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:msub>
                    <mml:mi>V</mml:mi>
                    <mml:mi>r</mml:mi>
                  </mml:msub>
                  <mml:msubsup>
                    <mml:mi>Σ</mml:mi>
                    <mml:mi>r</mml:mi>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mn>1</mml:mn>
                        <mml:mo>/</mml:mo>
                        <mml:mn>2</mml:mn>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msubsup>
                  <mml:mo>.</mml:mo>
                </mml:mtd>
              </mml:mtr>
            </mml:mtable>
          </mml:math>
        </disp-formula>
        <p>Here, <inline-formula><mml:math><mml:mrow><mml:mi> A </mml:mi><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math><mml:mrow><mml:mi> B </mml:mi><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> p </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> . Under this representation, the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mi> A </mml:mi></mml:math></inline-formula> characterizes the latent features of the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th depth sample, whereas the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mi> B </mml:mi></mml:math></inline-formula> describes how the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th logging variable is associated with the latent factors.</p>
        <p>This factorized form provides an effective way to represent the shared correlation structure among different well-log curves. Such a low-rank property is particularly advantageous for recovering missing entries in the data matrix, because it allows the reconstruction model to exploit the underlying dependencies among variables rather than treating each log independently [<xref ref-type="bibr" rid="B17">17</xref>] [<xref ref-type="bibr" rid="B18">18</xref>]. Low-rank matrix recovery has also been applied to oil-well logging data processing, showing its potential for extracting coherent logging responses and suppressing irregular perturbations in logging curves [<xref ref-type="bibr" rid="B19">19</xref>].</p>
      </sec>
      <sec id="sec2dot3">
        <title>2.3. Depth Continuity Prior</title>
        <p>In addition to the mutual correlation among different logging curves, a single well-log curve usually exhibits continuity along the depth direction within the same borehole. Under normal geological conditions, the measured response changes gradually with depth rather than fluctuating abruptly. Sharp variations generally occur near formation boundaries or within intervals with obvious lithological changes. Therefore, depthwise continuity can be regarded as an important geological prior for missing-log reconstruction [<xref ref-type="bibr" rid="B20">20</xref>] [<xref ref-type="bibr" rid="B21">21</xref>]. Previous geophysical studies have also shown that introducing smoothness constraints based on well-log priors can improve both the numerical stability and the geological plausibility of inversion results [<xref ref-type="bibr" rid="B22">22</xref>]. Related studies on multiple well-seismic ties further indicate that depth-consistent alignment and joint use of well information are important for improving the consistency of subsurface interpretation [<xref ref-type="bibr" rid="B23">23</xref>].</p>
        <p>Although low-rank modeling can capture the global correlations among different logging variables, relying only on inter-log correlation may still lead to unreasonable local fluctuations in the reconstructed curve, especially within continuous missing intervals. In other words, global correlation information alone is not sufficient to guarantee that the reconstructed DTC curve is smooth and geologically consistent along depth. It is therefore necessary to further introduce a depth-continuity constraint to preserve the local morphology of the logging curve.</p>
        <p>For a depth-ordered logging sequence, depth continuity can be expressed by penalizing the first-order differences between adjacent depth samples. This penalty discourages abrupt changes between neighboring reconstructed values and encourages the recovered curve to vary smoothly along depth. Let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> v </mml:mi></mml:mstyle></mml:math></inline-formula> denote the current reconstructed estimate of a missing logging sequence, and let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle></mml:math></inline-formula> denote the corrected sequence after imposing the depth-continuity constraint. The correction can be formulated as an optimization problem in which <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle></mml:math></inline-formula> is required to remain close to <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> v </mml:mi></mml:mstyle></mml:math></inline-formula> while satisfying a smoothness requirement.</p>
        <p>This correction can be interpreted using the proximal-operator framework. For a proper convex function <inline-formula><mml:math><mml:mrow><mml:mi> g </mml:mi><mml:mrow><mml:mo> ( </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> , the proximal operator of <inline-formula><mml:math><mml:mi> g </mml:mi></mml:math></inline-formula> at <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> v </mml:mi></mml:mstyle></mml:math></inline-formula> is defined as </p>
        <disp-formula id="FD5">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mrow>
                  <mml:mtext>prox</mml:mtext>
                </mml:mrow>
                <mml:mi>g</mml:mi>
              </mml:msub>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>v</mml:mi>
                </mml:mstyle>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>=</mml:mo>
              <mml:mtext>arg</mml:mtext>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>x</mml:mi>
                </mml:mstyle>
              </mml:munder>
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mrow>
                  <mml:mfrac>
                    <mml:mn>1</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>x</mml:mi>
                          </mml:mstyle>
                          <mml:mo>−</mml:mo>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>v</mml:mi>
                          </mml:mstyle>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:mi>g</mml:mi>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mstyle mathvariant="bold" mathsize="normal">
                      <mml:mi>x</mml:mi>
                    </mml:mstyle>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mo>}</mml:mo>
              </mml:mrow>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>In this definition, <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> v </mml:mi></mml:mstyle></mml:math></inline-formula> is the input vector before correction, <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle></mml:math></inline-formula> is the optimized output vector, and <inline-formula><mml:math><mml:mrow><mml:mi> g </mml:mi><mml:mrow><mml:mo> ( </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> represents the structural constraint imposed on the solution. In the present study, <inline-formula><mml:math><mml:mrow><mml:mi> g </mml:mi><mml:mrow><mml:mo> ( </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is associated with the depthwise smoothness of the reconstructed log. The first term keeps the corrected sequence close to the original reconstruction, whereas the second term enforces the desired smoothness along depth. Thus, the proximal formulation provides a convenient optimization form for the subsequent depth-continuity correction. </p>
      </sec>
    </sec>
    <sec id="sec3">
      <title>3. Methodology</title>
      <sec id="sec3dot1">
        <title>3.1. Low-Rank Matrix Recovery Model</title>
        <p>Using the standardized multi-log matrix <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> and the observation mask <inline-formula><mml:math><mml:mi> Ω </mml:mi></mml:math></inline-formula> defined in Section 2.1, the missing DTC reconstruction problem can be formulated as a matrix recovery task. Since different logging variables are controlled by a limited number of common geological and petrophysical factors, the complete multi-log matrix is expected to have an approximately low-rank structure. A natural formulation is therefore to seek a recovered matrix <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> with the lowest possible rank while keeping the observed entries consistent with the available measurements: </p>
        <disp-formula id="FD6">
          <mml:math>
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>M</mml:mi>
                </mml:mstyle>
              </mml:munder>
              <mml:mfrac>
                <mml:mn>1</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:mfrac>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:mi>Ω</mml:mi>
                      <mml:mo>⊙</mml:mo>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>Z</mml:mi>
                          </mml:mstyle>
                          <mml:mo>−</mml:mo>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>M</mml:mi>
                          </mml:mstyle>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mi>F</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:mtext>
                 
              </mml:mtext>
              <mml:mtext>rank</mml:mtext>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>M</mml:mi>
                </mml:mstyle>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mo> ⊙ </mml:mo></mml:math></inline-formula> denotes the Hadamard product, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mrow><mml:mrow><mml:mo> ‖ </mml:mo><mml:mrow><mml:mtext>   </mml:mtext><mml:mo> ⋅ </mml:mo><mml:mtext>   </mml:mtext></mml:mrow><mml:mo> ‖ </mml:mo></mml:mrow></mml:mrow><mml:mi> F </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the Frobenius norm, and <inline-formula><mml:math><mml:mrow><mml:mi> λ </mml:mi><mml:mo> &gt; </mml:mo><mml:mn> 0 </mml:mn></mml:mrow></mml:math></inline-formula> controls the strength of the low-rank constraint.</p>
        <p>However, directly minimizing the rank function is computationally difficult because <inline-formula><mml:math><mml:mrow><mml:mtext> rank </mml:mtext><mml:mrow><mml:mo> ( </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is non-convex and discontinuous. In matrix recovery problems, a common strategy is to replace the rank function with the nuclear norm, which is defined as the sum of the singular values of the matrix: </p>
        <disp-formula id="FD7">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mstyle mathvariant="bold" mathsize="normal">
                      <mml:mi>M</mml:mi>
                    </mml:mstyle>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mtext>*</mml:mtext>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:munder>
                <mml:mstyle mathsize="140%" displaystyle="true">
                  <mml:mo>∑</mml:mo>
                </mml:mstyle>
                <mml:mi>k</mml:mi>
              </mml:munder>
              <mml:mtext>
                 
              </mml:mtext>
              <mml:msub>
                <mml:mi>σ</mml:mi>
                <mml:mi>k</mml:mi>
              </mml:msub>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>M</mml:mi>
                </mml:mstyle>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> σ </mml:mi><mml:mi> k </mml:mi></mml:msub><mml:mrow><mml:mo> ( </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> denotes the <inline-formula><mml:math><mml:mi> k </mml:mi></mml:math></inline-formula> -th singular value of <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> . The nuclear norm is widely used as a convex surrogate for the rank function and encourages the recovered matrix to have a low-rank structure. Singular-value-thresholding-based methods provide a representative computational framework for nuclear-norm-regularized matrix completion [<xref ref-type="bibr" rid="B24">24</xref>]. This leads to the following nuclear-norm regularized recovery model: </p>
        <disp-formula id="FD8">
          <mml:math display="inline">
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>M</mml:mi>
                </mml:mstyle>
              </mml:munder>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:mfrac>
                        <mml:mn>1</mml:mn>
                        <mml:mn>2</mml:mn>
                      </mml:mfrac>
                      <mml:mi>Ω</mml:mi>
                      <mml:mo>⊙</mml:mo>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>Z</mml:mi>
                          </mml:mstyle>
                          <mml:mo>−</mml:mo>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>M</mml:mi>
                          </mml:mstyle>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mi>F</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:msub>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mstyle mathvariant="bold" mathsize="normal">
                      <mml:mi>M</mml:mi>
                    </mml:mstyle>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mtext>*</mml:mtext>
              </mml:msub>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>For computational implementation, the low-rank matrix <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> is further represented by the product of two low-dimensional factor matrices:</p>
        <disp-formula id="FD9">
          <mml:math>
            <mml:mrow>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>M</mml:mi>
              </mml:mstyle>
              <mml:mo>=</mml:mo>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>A</mml:mi>
              </mml:mstyle>
              <mml:msup>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>B</mml:mi>
                </mml:mstyle>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> n </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:mi> p </mml:mi><mml:mo> × </mml:mo><mml:mi> r </mml:mi></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> are latent factor matrices, and <inline-formula><mml:math><mml:mrow><mml:mi> r </mml:mi><mml:mo> ≪ </mml:mo><mml:mtext> min </mml:mtext><mml:mrow><mml:mo> ( </mml:mo><mml:mrow><mml:mi> n </mml:mi><mml:mo> , </mml:mo><mml:mi> p </mml:mi></mml:mrow><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is the prescribed rank. Under this factorized representation, the recovery model used in this study is written as </p>
        <disp-formula id="FD10">
          <label>(1)</label>
          <mml:math>
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:mstyle mathvariant="bold" mathsize="normal">
                    <mml:mi>A</mml:mi>
                  </mml:mstyle>
                  <mml:mo>,</mml:mo>
                  <mml:mstyle mathvariant="bold" mathsize="normal">
                    <mml:mi>B</mml:mi>
                  </mml:mstyle>
                </mml:mrow>
              </mml:munder>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:mi>Ω</mml:mi>
                      <mml:mo>⊙</mml:mo>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>Z</mml:mi>
                          </mml:mstyle>
                          <mml:mo>−</mml:mo>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>A</mml:mi>
                          </mml:mstyle>
                          <mml:msup>
                            <mml:mstyle mathvariant="bold" mathsize="normal">
                              <mml:mi>B</mml:mi>
                            </mml:mstyle>
                            <mml:mo>⊤</mml:mo>
                          </mml:msup>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mi>F</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>A</mml:mi>
                        </mml:mstyle>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mi>F</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>B</mml:mi>
                        </mml:mstyle>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mi>F</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>In Equation (1), the reconstruction error is evaluated only on the observed entries, so that the available measurements are preserved during model fitting. The regularization term constrains the magnitude of the latent factors and improves the numerical stability of the solution [<xref ref-type="bibr" rid="B25">25</xref>]. Compared with directly estimating all entries of <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> , which involves <inline-formula><mml:math><mml:mrow><mml:mi> n </mml:mi><mml:mi> p </mml:mi></mml:mrow></mml:math></inline-formula> unknowns, the factorized form only requires <inline-formula><mml:math><mml:mrow><mml:mi> r </mml:mi><mml:mrow><mml:mo> ( </mml:mo><mml:mrow><mml:mi> n </mml:mi><mml:mo> + </mml:mo><mml:mi> p </mml:mi></mml:mrow><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> unknowns. When <inline-formula><mml:math><mml:mi> r </mml:mi></mml:math></inline-formula> is much smaller than <inline-formula><mml:math><mml:mrow><mml:mtext> min </mml:mtext><mml:mrow><mml:mo> ( </mml:mo><mml:mrow><mml:mi> n </mml:mi><mml:mo> , </mml:mo><mml:mi> p </mml:mi></mml:mrow><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> , this representation substantially reduces the number of parameters and makes the recovery problem more tractable.</p>
        <p>After optimization, the reconstructed standardized matrix is obtained as </p>
        <disp-formula id="FD11">
          <mml:math>
            <mml:mrow>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>M</mml:mi>
              </mml:mstyle>
              <mml:mo>=</mml:mo>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>A</mml:mi>
              </mml:mstyle>
              <mml:msup>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>B</mml:mi>
                </mml:mstyle>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>The entries of <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> corresponding to the missing positions in the DTC column are then used as the preliminary reconstruction results. These values are subsequently refined by the depth smoothness correction introduced in the next subsection.</p>
      </sec>
      <sec id="sec3dot2">
        <title>3.2. Depth Smoothness Correction</title>
        <p>The low-rank recovery model effectively captures the global correlations among logging variables, but it does not explicitly impose vertical continuity along the depth direction. In real subsurface settings, most well-log responses vary gradually with depth, except near formation boundaries or intervals with sharp lithological changes [<xref ref-type="bibr" rid="B26">26</xref>]. Therefore, when low-rank reconstruction is used alone, the recovered curves may still exhibit unrealistic local oscillations, especially within masked intervals. To incorporate this geological prior into the reconstruction process, a depth-wise smoothness correction is introduced.</p>
        <p>For each well <inline-formula><mml:math><mml:mi> w </mml:mi></mml:math></inline-formula> , suppose that the reconstructed target log contains <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> n </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> depth samples ordered by depth. Let <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle><mml:mi> w </mml:mi></mml:msub><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:msub><mml:mi> n </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> denote the corrected target-log segment for well <inline-formula><mml:math><mml:mi> w </mml:mi></mml:math></inline-formula> , and let <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> m </mml:mi></mml:mstyle><mml:mi> w </mml:mi></mml:msub><mml:mo> ∈ </mml:mo><mml:msup><mml:mi> ℝ </mml:mi><mml:mrow><mml:msub><mml:mi> n </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> denote the current reconstructed segment extracted from the low-rank estimate <inline-formula><mml:math><mml:mrow><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle><mml:mo> = </mml:mo><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle><mml:msup><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle><mml:mo> ⊤ </mml:mo></mml:msup></mml:mrow></mml:math></inline-formula> . Define the first-order difference operator </p>
        <disp-formula id="FD12">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>D</mml:mi>
                </mml:mstyle>
                <mml:mi>w</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mrow>
                <mml:mo>[</mml:mo>
                <mml:mrow>
                  <mml:mtable>
                    <mml:mtr>
                      <mml:mtd>
                        <mml:mrow>
                          <mml:mo>−</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>1</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋯</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                    </mml:mtr>
                    <mml:mtr>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mrow>
                          <mml:mo>−</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>1</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋯</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                    </mml:mtr>
                    <mml:mtr>
                      <mml:mtd>
                        <mml:mo>⋮</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋮</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋮</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋱</mml:mo>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mo>⋮</mml:mo>
                      </mml:mtd>
                    </mml:mtr>
                    <mml:mtr>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>0</mml:mn>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mrow>
                          <mml:mo>−</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                      </mml:mtd>
                      <mml:mtd>
                        <mml:mn>1</mml:mn>
                      </mml:mtd>
                    </mml:mtr>
                  </mml:mtable>
                </mml:mrow>
                <mml:mo>]</mml:mo>
              </mml:mrow>
              <mml:mo>∈</mml:mo>
              <mml:msup>
                <mml:mi>ℝ</mml:mi>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>n</mml:mi>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                      <mml:mo>−</mml:mo>
                      <mml:mn>1</mml:mn>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                  <mml:mo>×</mml:mo>
                  <mml:msub>
                    <mml:mi>n</mml:mi>
                    <mml:mi>w</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Then, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> D </mml:mi></mml:mstyle><mml:mi> w </mml:mi></mml:msub><mml:msub><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> x </mml:mi></mml:mstyle><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> represents the first-order differences between adjacent depth samples. Therefore, the depth-wise smoothness of the reconstructed DTC sequence can be quantified by the sum of squared adjacent differences: </p>
        <disp-formula id="FD13">
          <mml:math>
            <mml:mrow>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>D</mml:mi>
                        </mml:mstyle>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                      <mml:msub>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>x</mml:mi>
                        </mml:mstyle>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>=</mml:mo>
              <mml:munderover>
                <mml:mstyle mathsize="140%" displaystyle="true">
                  <mml:mo>∑</mml:mo>
                </mml:mstyle>
                <mml:mrow>
                  <mml:mi>i</mml:mi>
                  <mml:mo>=</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>n</mml:mi>
                    <mml:mi>w</mml:mi>
                  </mml:msub>
                  <mml:mo>−</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
              </mml:munderover>
              <mml:msup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>x</mml:mi>
                        <mml:mrow>
                          <mml:mi>w</mml:mi>
                          <mml:mo>,</mml:mo>
                          <mml:mi>i</mml:mi>
                          <mml:mo>+</mml:mo>
                          <mml:mn>1</mml:mn>
                        </mml:mrow>
                      </mml:msub>
                      <mml:mo>−</mml:mo>
                      <mml:msub>
                        <mml:mi>x</mml:mi>
                        <mml:mrow>
                          <mml:mi>w</mml:mi>
                          <mml:mo>,</mml:mo>
                          <mml:mi>i</mml:mi>
                        </mml:mrow>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>In the alternating optimization procedure, the following expression is treated as the smoothing subproblem within the overall objective. When updating the smoothing variable <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> x </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , the low-rank factors <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> and <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> , and hence the current estimate <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> m </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , are temporarily held fixed. Thus, <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> x </mml:mi><mml:mi> w </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the only unknown optimized in this step; the smoothing subproblem is not independent of the overall objective.</p>
        <p>To incorporate depthwise continuity while preserving consistency with the current low-rank reconstruction, a well-wise quadratic optimization subproblem is formulated as follows: </p>
        <disp-formula id="FD14">
          <label>(2)</label>
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>x</mml:mi>
                <mml:mi>w</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mtext>arg</mml:mtext>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mi>x</mml:mi>
              </mml:munder>
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mrow>
                  <mml:mfrac>
                    <mml:mn>1</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:mi>x</mml:mi>
                          <mml:mo>−</mml:mo>
                          <mml:msub>
                            <mml:mi>m</mml:mi>
                            <mml:mi>w</mml:mi>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:mfrac>
                    <mml:mi>α</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>D</mml:mi>
                            <mml:mi>w</mml:mi>
                          </mml:msub>
                          <mml:mi>x</mml:mi>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>}</mml:mo>
              </mml:mrow>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>This problem can be interpreted as a proximal step associated with the smoothness regularizer. Specifically, </p>
        <disp-formula id="FD15">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>x</mml:mi>
                <mml:mi>w</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:msub>
                <mml:mrow>
                  <mml:mtext>prox</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:mfrac>
                    <mml:mi>α</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>D</mml:mi>
                            <mml:mi>w</mml:mi>
                          </mml:msub>
                          <mml:mo>⋅</mml:mo>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
              </mml:msub>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>m</mml:mi>
                    <mml:mi>w</mml:mi>
                  </mml:msub>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where prox denotes the proximal operator. Since Equation (2) is a strictly convex quadratic problem, its solution is given in closed form by </p>
        <disp-formula id="FD16">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>x</mml:mi>
                <mml:mi>w</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:msup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:mi>I</mml:mi>
                      <mml:mo>+</mml:mo>
                      <mml:mi>α</mml:mi>
                      <mml:msubsup>
                        <mml:mi>D</mml:mi>
                        <mml:mi>w</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                      <mml:msub>
                        <mml:mi>D</mml:mi>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mrow>
                  <mml:mo>−</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
              </mml:msup>
              <mml:msub>
                <mml:mi>m</mml:mi>
                <mml:mi>w</mml:mi>
              </mml:msub>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Although the complete reconstruction process can be conceptually associated with the following regularized objective, the optimization in this study is implemented in a decoupled manner. Specifically, low-rank recovery is first performed through ALS by alternately updating <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> and <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> using the observed entries. With the resulting low-rank estimate held fixed, the depth-wise smoothness term is then solved as the well-wise proximal subproblem above. The smoothed DTC estimates are used to form the final reconstruction and are not fed back into a subsequent ALS factor update. Accordingly, the overall reconstruction model can be written as </p>
        <disp-formula id="FD17">
          <label>(3)</label>
          <mml:math>
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:mi>A</mml:mi>
                  <mml:mo>,</mml:mo>
                  <mml:mi>B</mml:mi>
                  <mml:mo>,</mml:mo>
                  <mml:mrow>
                    <mml:mo>{</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>x</mml:mi>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>}</mml:mo>
                  </mml:mrow>
                </mml:mrow>
              </mml:munder>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:mi>Ω</mml:mi>
                      <mml:mo>⊙</mml:mo>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mrow>
                          <mml:mi>Z</mml:mi>
                          <mml:mo>−</mml:mo>
                          <mml:mi>A</mml:mi>
                          <mml:msup>
                            <mml:mi>B</mml:mi>
                            <mml:mo>⊤</mml:mo>
                          </mml:msup>
                        </mml:mrow>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mi>F</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mi>A</mml:mi>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mi>F</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mi>B</mml:mi>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mi>F</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>+</mml:mo>
              <mml:mi>α</mml:mi>
              <mml:munder>
                <mml:mstyle mathsize="140%" displaystyle="true">
                  <mml:mo>∑</mml:mo>
                </mml:mstyle>
                <mml:mi>w</mml:mi>
              </mml:munder>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>D</mml:mi>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                      <mml:msub>
                        <mml:mi>x</mml:mi>
                        <mml:mi>w</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>In practice, Equation (3) is not solved as a fully coupled joint update. The implementation first completes the ALS low-rank fitting using only the observed entries, constructs the preliminary estimate, and then solves the smoothing subproblem with the low-rank factors fixed. The resulting smoothed values are not reintroduced into the ALS factor updates. Therefore, the depth-wise correction is a subproblem-level realization of the smoothness regularization term rather than an additional ALS update. </p>
      </sec>
      <sec id="sec3dot3">
        <title>3.3. Optimization Procedure</title>
        <p>The reconstruction model presented here is solved via an alternating least squares (ALS) procedure, a choice motivated by the bilinear structure of the low-rank factorization problem in <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> and <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> : a joint optimization of both factor matrices proves challenging, whereas fixing one of them reduces the subproblem to a regularized least-squares formulation that admits a closed-form solution [<xref ref-type="bibr" rid="B27">27</xref>] [<xref ref-type="bibr" rid="B28">28</xref>]. The overall optimization therefore proceeds by alternately updating <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> and <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> .</p>
        <p>Let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> stand for the standardized complete data matrix, and let <inline-formula><mml:math display="inline"><mml:mi> Ω </mml:mi></mml:math></inline-formula> represent the observation mask matrix. An entry <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> set to 1 indicates that the corresponding value in <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> is treated as observed throughout the optimization. An entry set to 0 signals that the value is regarded as missing and therefore becomes a target for reconstruction. In the masking experiment, the true values at the masked positions do remain present within <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> . Through the action of the mask <inline-formula><mml:math display="inline"><mml:mi> Ω </mml:mi></mml:math></inline-formula> , however, they are kept apart from the ALS fitting process and play no role in guiding the updates.</p>
        <p>The optimization is initialized by assigning random values to the factor matrices <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> and <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> . The two factor matrices are then updated iteratively using the ALS scheme, which is commonly adopted in low-rank matrix completion problems [<xref ref-type="bibr" rid="B29">29</xref>]. In each iteration, one factor matrix is fixed while the other is updated by solving a set of regularized least-squares subproblems. The detailed update procedure is described as follows. </p>
        <p>Step 1. Update of <italic>A</italic></p>
        <p>When <inline-formula><mml:math><mml:mi> B </mml:mi></mml:math></inline-formula> is fixed, the update of <inline-formula><mml:math><mml:mi> A </mml:mi></mml:math></inline-formula> can be performed row by row. For the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th sample, only the observed entries indicated by the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mtext> Ω </mml:mtext></mml:math></inline-formula> are used. Let</p>
        <disp-formula id="FD18">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi mathvariant="script">O</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mo>
              </mml:mo>
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mrow>
                  <mml:mi>j</mml:mi>
                  <mml:mo>|</mml:mo>
                  <mml:msub>
                    <mml:mi>Ω</mml:mi>
                    <mml:mrow>
                      <mml:mi>i</mml:mi>
                      <mml:mi>j</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>=</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
                <mml:mo>}</mml:mo>
              </mml:mrow>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Accordingly, the corresponding submatrix of <inline-formula><mml:math><mml:mi> B </mml:mi></mml:math></inline-formula> is denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> B </mml:mi><mml:mi> i </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , and the observed values in the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mi> Z </mml:mi></mml:math></inline-formula> are denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Z </mml:mi><mml:mi> i </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> . The update of the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mi> A </mml:mi></mml:math></inline-formula> , denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> A </mml:mi><mml:mi> i </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , is obtained by solving the following regularized least-squares problem:</p>
        <disp-formula id="FD19">
          <mml:math>
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>A</mml:mi>
                    <mml:mi>i</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:munder>
              <mml:mfrac>
                <mml:mn>1</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:mfrac>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>Z</mml:mi>
                        <mml:mi>i</mml:mi>
                      </mml:msub>
                      <mml:mo>−</mml:mo>
                      <mml:msub>
                        <mml:mi>A</mml:mi>
                        <mml:mi>i</mml:mi>
                      </mml:msub>
                      <mml:msubsup>
                        <mml:mi>B</mml:mi>
                        <mml:mi>i</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mfrac>
                <mml:mi>λ</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:mfrac>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>A</mml:mi>
                        <mml:mi>i</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Taking the derivative with respect to <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> A </mml:mi><mml:mi> i </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> yields </p>
        <disp-formula id="FD20">
          <mml:math>
            <mml:mrow>
              <mml:mfrac>
                <mml:mo>∂</mml:mo>
                <mml:mrow>
                  <mml:mo>∂</mml:mo>
                  <mml:msub>
                    <mml:mi>A</mml:mi>
                    <mml:mi>i</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:mfrac>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:mfrac>
                    <mml:mn>1</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>Z</mml:mi>
                            <mml:mi>i</mml:mi>
                          </mml:msub>
                          <mml:mo>−</mml:mo>
                          <mml:msub>
                            <mml:mi>A</mml:mi>
                            <mml:mi>i</mml:mi>
                          </mml:msub>
                          <mml:msubsup>
                            <mml:mi>B</mml:mi>
                            <mml:mi>i</mml:mi>
                            <mml:mo>⊤</mml:mo>
                          </mml:msubsup>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:mfrac>
                    <mml:mi>λ</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>A</mml:mi>
                            <mml:mi>i</mml:mi>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>=</mml:mo>
              <mml:mo>−</mml:mo>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>Z</mml:mi>
                    <mml:mi>i</mml:mi>
                  </mml:msub>
                  <mml:mo>−</mml:mo>
                  <mml:msub>
                    <mml:mi>A</mml:mi>
                    <mml:mi>i</mml:mi>
                  </mml:msub>
                  <mml:msubsup>
                    <mml:mi>B</mml:mi>
                    <mml:mi>i</mml:mi>
                    <mml:mo>⊤</mml:mo>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:msub>
                <mml:mi>B</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:msub>
                <mml:mi>A</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Setting the derivative to zero gives the corresponding normal equation. Therefore, the closed-form update of <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> A </mml:mi><mml:mi> i </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is </p>
        <disp-formula id="FD21">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>A</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:msub>
                <mml:mi>Z</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:msub>
                <mml:mi>B</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:msup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:msubsup>
                        <mml:mi>B</mml:mi>
                        <mml:mi>i</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mi>i</mml:mi>
                      </mml:msub>
                      <mml:mo>+</mml:mo>
                      <mml:mi>λ</mml:mi>
                      <mml:mi>I</mml:mi>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mrow>
                  <mml:mo>−</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>This step estimates the latent representation of the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th sample based only on its available observations.</p>
        <p>Step 2. Update of <italic>B</italic></p>
        <p>Similarly, when <inline-formula><mml:math><mml:mi> A </mml:mi></mml:math></inline-formula> is fixed, the update of <inline-formula><mml:math><mml:mi> B </mml:mi></mml:math></inline-formula> can be performed row by row. For the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th logging variable, only the observed entries indicated by the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th column of Ω are used. Let</p>
        <disp-formula id="FD22">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi mathvariant="script">O</mml:mi>
                <mml:mi>j</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mrow>
                  <mml:mi>i</mml:mi>
                  <mml:mo>|</mml:mo>
                  <mml:msub>
                    <mml:mi>Ω</mml:mi>
                    <mml:mrow>
                      <mml:mi>i</mml:mi>
                      <mml:mi>j</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>=</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
                <mml:mo>}</mml:mo>
              </mml:mrow>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Then, the corresponding submatrix of <inline-formula><mml:math><mml:mi> A </mml:mi></mml:math></inline-formula> is denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> A </mml:mi><mml:mi> j </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , and the observed values in the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th column of <inline-formula><mml:math><mml:mi> Z </mml:mi></mml:math></inline-formula> are denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Z </mml:mi><mml:mi> j </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> . The update of the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th row of <inline-formula><mml:math><mml:mi> B </mml:mi></mml:math></inline-formula> , denoted by <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> B </mml:mi><mml:mi> j </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> , is obtained by solving the following regularized least-squares problem: </p>
        <disp-formula id="FD23">
          <mml:math>
            <mml:mrow>
              <mml:munder>
                <mml:mrow>
                  <mml:mtext>min</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>B</mml:mi>
                    <mml:mi>j</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:munder>
              <mml:mfrac>
                <mml:mn>1</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:mfrac>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>Z</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                      <mml:mo>−</mml:mo>
                      <mml:msub>
                        <mml:mi>A</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                      <mml:msubsup>
                        <mml:mi>B</mml:mi>
                        <mml:mi>j</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>+</mml:mo>
              <mml:mfrac>
                <mml:mi>λ</mml:mi>
                <mml:mn>2</mml:mn>
              </mml:mfrac>
              <mml:msubsup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>‖</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>B</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                    </mml:mrow>
                    <mml:mo>‖</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mn>2</mml:mn>
                <mml:mn>2</mml:mn>
              </mml:msubsup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Taking the derivative with respect to <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> B </mml:mi><mml:mi> j </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> yields</p>
        <disp-formula id="FD24">
          <mml:math>
            <mml:mrow>
              <mml:mfrac>
                <mml:mo>∂</mml:mo>
                <mml:mrow>
                  <mml:mo>∂</mml:mo>
                  <mml:msub>
                    <mml:mi>B</mml:mi>
                    <mml:mi>j</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:mfrac>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:mfrac>
                    <mml:mn>1</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>Z</mml:mi>
                            <mml:mi>j</mml:mi>
                          </mml:msub>
                          <mml:mo>−</mml:mo>
                          <mml:msub>
                            <mml:mi>A</mml:mi>
                            <mml:mi>j</mml:mi>
                          </mml:msub>
                          <mml:msubsup>
                            <mml:mi>B</mml:mi>
                            <mml:mi>j</mml:mi>
                            <mml:mo>⊤</mml:mo>
                          </mml:msubsup>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                  <mml:mo>+</mml:mo>
                  <mml:mfrac>
                    <mml:mi>λ</mml:mi>
                    <mml:mn>2</mml:mn>
                  </mml:mfrac>
                  <mml:msubsup>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>‖</mml:mo>
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>B</mml:mi>
                            <mml:mi>j</mml:mi>
                          </mml:msub>
                        </mml:mrow>
                        <mml:mo>‖</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:mn>2</mml:mn>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>=</mml:mo>
              <mml:mo>−</mml:mo>
              <mml:msup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:msub>
                        <mml:mi>Z</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                      <mml:mo>−</mml:mo>
                      <mml:msub>
                        <mml:mi>A</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                      <mml:msubsup>
                        <mml:mi>B</mml:mi>
                        <mml:mi>j</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:msub>
                <mml:mi>A</mml:mi>
                <mml:mi>j</mml:mi>
              </mml:msub>
              <mml:mo>+</mml:mo>
              <mml:mi>λ</mml:mi>
              <mml:msub>
                <mml:mi>B</mml:mi>
                <mml:mi>j</mml:mi>
              </mml:msub>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>Setting the derivative to zero gives the corresponding normal equation. Therefore, the closed-form update of <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> B </mml:mi><mml:mi> j </mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is </p>
        <disp-formula id="FD25">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>B</mml:mi>
                <mml:mi>j</mml:mi>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:msubsup>
                <mml:mi>Z</mml:mi>
                <mml:mi>j</mml:mi>
                <mml:mo>⊤</mml:mo>
              </mml:msubsup>
              <mml:msub>
                <mml:mi>A</mml:mi>
                <mml:mi>j</mml:mi>
              </mml:msub>
              <mml:msup>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mo>(</mml:mo>
                    <mml:mrow>
                      <mml:msubsup>
                        <mml:mi>A</mml:mi>
                        <mml:mi>j</mml:mi>
                        <mml:mo>⊤</mml:mo>
                      </mml:msubsup>
                      <mml:msub>
                        <mml:mi>A</mml:mi>
                        <mml:mi>j</mml:mi>
                      </mml:msub>
                      <mml:mo>+</mml:mo>
                      <mml:mi>λ</mml:mi>
                      <mml:mi>I</mml:mi>
                    </mml:mrow>
                    <mml:mo>)</mml:mo>
                  </mml:mrow>
                </mml:mrow>
                <mml:mrow>
                  <mml:mo>−</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>This step estimates the latent representation of the <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> -th logging variable under the current sample-factor matrix. </p>
        <p>Step 3. Construction of the Low-Rank Estimate</p>
        <p>After both factor matrices have been updated, the low-rank estimate of the standardized data matrix is obtained as </p>
        <disp-formula id="FD26">
          <mml:math>
            <mml:mrow>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>M</mml:mi>
              </mml:mstyle>
              <mml:mo>=</mml:mo>
              <mml:mstyle mathvariant="bold" mathsize="normal">
                <mml:mi>A</mml:mi>
              </mml:mstyle>
              <mml:msup>
                <mml:mstyle mathvariant="bold" mathsize="normal">
                  <mml:mi>B</mml:mi>
                </mml:mstyle>
                <mml:mo>⊤</mml:mo>
              </mml:msup>
              <mml:mo>.</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>The matrix <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle></mml:math></inline-formula> provides model-based estimates for all entries of the standardized data matrix. At this stage, however, it is only regarded as an intermediate estimate. The final reconstruction is obtained after the subsequent depth-wise smoothing and data-consistency enforcement steps.</p>
        <p>Step 4. Depth Smoothness Correction</p>
        <p>For each well, the DTC sequence extracted from the preliminary estimate <inline-formula><mml:math><mml:mrow><mml:msup><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> M </mml:mi></mml:mstyle><mml:mrow><mml:mrow><mml:mo> ( </mml:mo><mml:mn> 0 </mml:mn><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> is refined using the depth-wise smoothing strategy. This step reduces unrealistic local fluctuations and improves the continuity of the reconstructed DTC curve along depth. The resulting smoothed DTC values are used for final data completion only; they are not used to update <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> A </mml:mi></mml:mstyle></mml:math></inline-formula> or <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> B </mml:mi></mml:mstyle></mml:math></inline-formula> in a later ALS iteration. Let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mover accent="true"><mml:mi> M </mml:mi><mml:mo> ˜ </mml:mo></mml:mover></mml:mstyle></mml:math></inline-formula> denote the matrix after this smoothing correction. </p>
        <p>Step 5. Restoration of Observed Entries</p>
        <p>After the depth-wise smoothing step, data consistency is enforced by combining the original observed entries with the smoothed reconstructed values at the masked positions. Let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mover accent="true"><mml:mi> M </mml:mi><mml:mo> ˜ </mml:mo></mml:mover></mml:mstyle></mml:math></inline-formula> denote the matrix after low-rank reconstruction and depth-wise smoothing, and let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mover accent="true"><mml:mi> M </mml:mi><mml:mo> ^ </mml:mo></mml:mover></mml:mstyle></mml:math></inline-formula> denote the final standardized reconstruction. Then, </p>
        <disp-formula id="FD27">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mover accent="true">
                  <mml:mi>M</mml:mi>
                  <mml:mo>^</mml:mo>
                </mml:mover>
                <mml:mrow>
                  <mml:mi>i</mml:mi>
                  <mml:mi>j</mml:mi>
                </mml:mrow>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mrow>
                <mml:mo>{</mml:mo>
                <mml:mrow>
                  <mml:mtable columnalign="left">
                    <mml:mtr columnalign="left">
                      <mml:mtd columnalign="left">
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>Z</mml:mi>
                            <mml:mrow>
                              <mml:mi>i</mml:mi>
                              <mml:mi>j</mml:mi>
                            </mml:mrow>
                          </mml:msub>
                          <mml:mo>,</mml:mo>
                        </mml:mrow>
                      </mml:mtd>
                      <mml:mtd columnalign="left">
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>Ω</mml:mi>
                            <mml:mrow>
                              <mml:mi>i</mml:mi>
                              <mml:mi>j</mml:mi>
                            </mml:mrow>
                          </mml:msub>
                          <mml:mo>=</mml:mo>
                          <mml:mn>1</mml:mn>
                          <mml:mo>,</mml:mo>
                        </mml:mrow>
                      </mml:mtd>
                    </mml:mtr>
                    <mml:mtr columnalign="left">
                      <mml:mtd columnalign="left">
                        <mml:mrow>
                          <mml:msub>
                            <mml:mover accent="true">
                              <mml:mi>M</mml:mi>
                              <mml:mo>˜</mml:mo>
                            </mml:mover>
                            <mml:mrow>
                              <mml:mi>i</mml:mi>
                              <mml:mi>j</mml:mi>
                            </mml:mrow>
                          </mml:msub>
                          <mml:mo>,</mml:mo>
                        </mml:mrow>
                      </mml:mtd>
                      <mml:mtd columnalign="left">
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>Ω</mml:mi>
                            <mml:mrow>
                              <mml:mi>i</mml:mi>
                              <mml:mi>j</mml:mi>
                            </mml:mrow>
                          </mml:msub>
                          <mml:mo>=</mml:mo>
                          <mml:mn>0.</mml:mn>
                        </mml:mrow>
                      </mml:mtd>
                    </mml:mtr>
                  </mml:mtable>
                </mml:mrow>
              </mml:mrow>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>This step does not indicate that the observed entries are reconstructed. Rather, it ensures that the known measurements remain fixed in the final output, whereas only the masked entries are filled by the reconstruction model.</p>
        <p>Step 6. Iteration and Inverse Transformation</p>
        <p>The ALS factor updates are repeated until the preset maximum number of iterations is reached. After the final low-rank estimate has been obtained, the depth-wise smoothing and data-consistency steps are applied once to produce the final standardized reconstruction <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mover accent="true"><mml:mi> M </mml:mi><mml:mo> ^ </mml:mo></mml:mover></mml:mstyle></mml:math></inline-formula> . The final standardized reconstruction is then converted back to the original measurement scale using the inverse standardization parameters obtained during preprocessing. For the RT variable, an additional inverse transformation is required because a logarithmic transformation was applied before standardization. Therefore, the corresponding values are exponentiated after inverse standardization to recover the original resistivity scale. Finally, the reconstructed DTC values are extracted from the DTC column for subsequent evaluation and visualization.</p>
      </sec>
      <sec id="sec3dot4">
        <title>3.4. Algorithm Framework</title>
        <fig id="fig1">
          <label>Figure 1</label>
          <caption>
            <p>Framework of matrix recovery algorithm.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId320.jpeg?20261118013136" />
        </fig>
        <table-wrap id="tbl1">
          <label>Table 1</label>
          <table>
            <tbody>
              <tr>
                <td>
                  <bold>Algorithm 1</bold>
                  DTC reconstruction framework based on low-rank matrix factorization and depth-wise smoothness correction
                </td>
              </tr>
              <tr>
                <td>
                  <bold>Input:</bold>
                  standardized multi-log matrix
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>Z</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  ; observation mask matrix
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>Ω</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  ; well index information and depth order; rank
                  <inline-formula>
                    <mml:math>
                      <mml:mi>r</mml:mi>
                    </mml:math>
                  </inline-formula>
                  ; regularization coefficient
                  <inline-formula>
                    <mml:math>
                      <mml:mi>λ</mml:mi>
                    </mml:math>
                  </inline-formula>
                  ; smoothness coefficient
                  <inline-formula>
                    <mml:math>
                      <mml:mi>α</mml:mi>
                    </mml:math>
                  </inline-formula>
                  ; maximum number of iterations
                  <inline-formula>
                    <mml:math>
                      <mml:mi>T</mml:mi>
                    </mml:math>
                  </inline-formula>
                  <bold>Output:</bold>
                  reconstructed DTC log in the original measurement scale1: Initialize the factor matrices
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>A</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  and
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>B</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  with small random values2:
                  <bold>for</bold>
                  <inline-formula>
                    <mml:math>
                      <mml:mrow>
                        <mml:mi>t</mml:mi>
                        <mml:mo>=</mml:mo>
                        <mml:mn>1</mml:mn>
                      </mml:mrow>
                    </mml:math>
                  </inline-formula>
                  to
                  <inline-formula>
                    <mml:math>
                      <mml:mi>T</mml:mi>
                    </mml:math>
                  </inline-formula>
                  do3: Update the sample-factor matrix
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>A</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  row by row using only the observed entries indicated by
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>Ω</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  4: Update the variable-factor matrix
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>B</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  row by row using only the observed entries indicated by
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>Ω</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  5: Construct the preliminary low-rank estimate:
                  <inline-formula>
                    <mml:math>
                      <mml:mrow>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>M</mml:mi>
                        </mml:mstyle>
                        <mml:mo>=</mml:mo>
                        <mml:mstyle mathvariant="bold" mathsize="normal">
                          <mml:mi>A</mml:mi>
                        </mml:mstyle>
                        <mml:msup>
                          <mml:mstyle mathvariant="bold" mathsize="normal">
                            <mml:mi>B</mml:mi>
                          </mml:mstyle>
                          <mml:mo>⊤</mml:mo>
                        </mml:msup>
                      </mml:mrow>
                    </mml:math>
                  </inline-formula>
                  6: The preliminary estimate
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>M</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  is used only for the ALS iteration; smoothing is not fed back into the factor updates7:
                  <bold>end</bold>
                  8: After the ALS iterations, extract the DTC sequence of each well from
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mi>M</mml:mi>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  and apply depth-wise smoothness correction based on the first-order difference operator, yielding
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mover accent="true">
                          <mml:mi>M</mml:mi>
                          <mml:mo>˜</mml:mo>
                        </mml:mover>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  9: Enforce data consistency by restoring the observed entries:
                  <inline-formula>
                    <mml:math>
                      <mml:mrow>
                        <mml:msub>
                          <mml:mstyle mathsize="normal" mathvariant="bold">
                            <mml:mover accent="true">
                              <mml:mi>M</mml:mi>
                              <mml:mo>^</mml:mo>
                            </mml:mover>
                          </mml:mstyle>
                          <mml:mrow>
                            <mml:mi>i</mml:mi>
                            <mml:mi>j</mml:mi>
                          </mml:mrow>
                        </mml:msub>
                        <mml:mo>=</mml:mo>
                        <mml:mrow>
                          <mml:mo>{</mml:mo>
                          <mml:mrow>
                            <mml:mtable columnalign="left">
                              <mml:mtr columnalign="left">
                                <mml:mtd columnalign="left">
                                  <mml:mrow>
                                    <mml:msub>
                                      <mml:mstyle mathsize="normal" mathvariant="bold">
                                        <mml:mi>Z</mml:mi>
                                      </mml:mstyle>
                                      <mml:mrow>
                                        <mml:mi>i</mml:mi>
                                        <mml:mi>j</mml:mi>
                                      </mml:mrow>
                                    </mml:msub>
                                    <mml:mo>,</mml:mo>
                                  </mml:mrow>
                                </mml:mtd>
                                <mml:mtd columnalign="left">
                                  <mml:mrow>
                                    <mml:msub>
                                      <mml:mtext>Ω</mml:mtext>
                                      <mml:mrow>
                                        <mml:mi>i</mml:mi>
                                        <mml:mi>j</mml:mi>
                                      </mml:mrow>
                                    </mml:msub>
                                    <mml:mo>=</mml:mo>
                                    <mml:mn>1</mml:mn>
                                    <mml:mo>,</mml:mo>
                                  </mml:mrow>
                                </mml:mtd>
                              </mml:mtr>
                              <mml:mtr columnalign="left">
                                <mml:mtd columnalign="left">
                                  <mml:mrow>
                                    <mml:msub>
                                      <mml:mstyle mathsize="normal" mathvariant="bold">
                                        <mml:mover accent="true">
                                          <mml:mi>M</mml:mi>
                                          <mml:mo>˜</mml:mo>
                                        </mml:mover>
                                      </mml:mstyle>
                                      <mml:mrow>
                                        <mml:mi>i</mml:mi>
                                        <mml:mi>j</mml:mi>
                                      </mml:mrow>
                                    </mml:msub>
                                    <mml:mo>,</mml:mo>
                                  </mml:mrow>
                                </mml:mtd>
                                <mml:mtd columnalign="left">
                                  <mml:mrow>
                                    <mml:msub>
                                      <mml:mtext>Ω</mml:mtext>
                                      <mml:mrow>
                                        <mml:mi>i</mml:mi>
                                        <mml:mi>j</mml:mi>
                                      </mml:mrow>
                                    </mml:msub>
                                    <mml:mo>=</mml:mo>
                                    <mml:mn>0.</mml:mn>
                                  </mml:mrow>
                                </mml:mtd>
                              </mml:mtr>
                            </mml:mtable>
                          </mml:mrow>
                        </mml:mrow>
                      </mml:mrow>
                    </mml:math>
                  </inline-formula>
                  10: Apply inverse standardization to
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mover accent="true">
                          <mml:mi>M</mml:mi>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  to obtain the reconstructed matrix
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mover accent="true">
                          <mml:mi>X</mml:mi>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  in the original measurement scale11: Extract the reconstructed DTC values from
                  <inline-formula>
                    <mml:math>
                      <mml:mstyle mathvariant="bold" mathsize="normal">
                        <mml:mover accent="true">
                          <mml:mi>X</mml:mi>
                          <mml:mo>^</mml:mo>
                        </mml:mover>
                      </mml:mstyle>
                    </mml:math>
                  </inline-formula>
                  for quantitative evaluation, visualization, and comparison with baseline methods
                </td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p>To clearly describe the implementation procedure of the proposed reconstruction method, the main computational steps are summarized in Algorithm 1. Algorithm 1 presents the iterative low-rank factor update, depth-wise smoothness correction, data-consistency restoration, and inverse transformation steps in a pseudo-code form. In addition, <xref ref-type="fig" rid="fig1">Figure 1</xref> provides a graphical overview of the reconstruction workflow, illustrating the relationship among the input data, iterative matrix recovery process, and final reconstructed DTC output.</p>
        <p>For a more intuitive understanding of the proposed framework, the graphical workflow of the reconstruction procedure is shown in <xref ref-type="fig" rid="fig1">Figure 1</xref>.</p>
      </sec>
    </sec>
    <sec id="sec4">
      <title>4. Experimental Setup</title>
      <sec id="sec4dot1">
        <title>4.1. Dataset and Preprocessing</title>
        <p>The dataset used in this study was supplied as a complete multi-log spreadsheet containing ten wells labeled W01 - W10. Each well contains 600 samples over a depth range of 2500.0 - 2799.5 m, with a uniform depth-grid spacing of 0.5 m, yielding 6000 samples in total. Records were included in the analysis on the basis of complete records for DTC and all six auxiliary variables. Before artificial masking, all selected wells therefore contain complete DTC curves, and the original DTC values serve as reference values in the artificial-masking experiments. During reconstruction, data from all wells are organized into a unified multi-log matrix for joint multi-well reconstruction. The main information of the dataset is summarized in <bold>Table 1</bold>.</p>
        <p><bold>Table 1.</bold> Summary of the dataset used in this study.</p>
        <table-wrap id="tbl2">
          <label>Table 2</label>
          <table>
            <tbody>
              <tr>
                <td>Item</td>
                <td>Value</td>
              </tr>
              <tr>
                <td>Number of wells</td>
                <td>10</td>
              </tr>
              <tr>
                <td>Total number of valid samples</td>
                <td>6000</td>
              </tr>
              <tr>
                <td>Samples per well</td>
                <td>600</td>
              </tr>
              <tr>
                <td>Depth range</td>
                <td>2500.0 - 2799.5 m</td>
              </tr>
              <tr>
                <td>Synthetic depth-grid spacing</td>
                <td>red0.5 m</td>
              </tr>
              <tr>
                <td>Target log</td>
                <td>DTC</td>
              </tr>
              <tr>
                <td>Auxiliary logs</td>
                <td>GR, RHOB, NPHI, RT, PHIT, SW</td>
              </tr>
              <tr>
                <td>Reconstruction mode</td>
                <td>Joint multi-well reconstruction</td>
              </tr>
              <tr>
                <td>Evaluation mode</td>
                <td>Well-wise masked-sample evaluation</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <p>PHIT and SW were imported directly from the supplied dataset rather than calculated within the reconstruction code. However, the supplied materials do not document the original data source, geological setting, initial well-selection procedure, or the methods used to obtain PHIT and SW. Consequently, whether these auxiliary variables were originally derived from DTC cannot be established. If either variable depends directly or indirectly on DTC, retaining its original values at masked DTC locations may introduce target-information leakage. The present results should therefore be interpreted as reconstruction performance conditional on the supplied auxiliary variables, rather than as evidence of performance using independently obtained inputs.</p>
        <p>Before reconstruction, preprocessing is performed to improve the numerical stability of the optimization. Since the RT curve usually exhibits a highly skewed distribution and may contain values with large magnitudes, a logarithmic transformation is first applied to RT: </p>
        <disp-formula id="FD28">
          <mml:math>
            <mml:mrow>
              <mml:msubsup>
                <mml:mi>X</mml:mi>
                <mml:mrow>
                  <mml:mi>i</mml:mi>
                  <mml:mo>,</mml:mo>
                  <mml:mtext>RT</mml:mtext>
                </mml:mrow>
                <mml:mrow>
                  <mml:mtext>log</mml:mtext>
                </mml:mrow>
              </mml:msubsup>
              <mml:mo>=</mml:mo>
              <mml:mtext>log</mml:mtext>
              <mml:mrow>
                <mml:mo>(</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>X</mml:mi>
                    <mml:mrow>
                      <mml:mi>i</mml:mi>
                      <mml:mo>,</mml:mo>
                      <mml:mtext>RT</mml:mtext>
                    </mml:mrow>
                  </mml:msub>
                </mml:mrow>
                <mml:mo>)</mml:mo>
              </mml:mrow>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> X </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mo> , </mml:mo><mml:mtext> RT </mml:mtext></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> denotes the original RT value at the <inline-formula><mml:math><mml:mi> i </mml:mi></mml:math></inline-formula> -th depth sample, and <inline-formula><mml:math><mml:mrow><mml:msubsup><mml:mi> X </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mo> , </mml:mo><mml:mtext> RT </mml:mtext></mml:mrow><mml:mrow><mml:mtext> log </mml:mtext></mml:mrow></mml:msubsup></mml:mrow></mml:math></inline-formula> denotes the corresponding log-transformed value. This transformation reduces the influence of extreme resistivity values.</p>
        <p>For each masking experiment, the masking pattern is generated before standardization. Let <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 1 </mml:mn></mml:mrow></mml:math></inline-formula> indicate that entry <inline-formula><mml:math><mml:mrow><mml:mrow><mml:mo> ( </mml:mo><mml:mrow><mml:mi> i </mml:mi><mml:mo> , </mml:mo><mml:mi> j </mml:mi></mml:mrow><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is available to the model and <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 0 </mml:mn></mml:mrow></mml:math></inline-formula> indicate that it is masked. The mean and standard deviation for each logging variable are calculated using only its unmasked observations. Let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> X </mml:mi></mml:mstyle></mml:math></inline-formula> denote the preprocessed data matrix before standardization, and let <inline-formula><mml:math><mml:mstyle mathvariant="bold" mathsize="normal"><mml:mi> Z </mml:mi></mml:mstyle></mml:math></inline-formula> denote the standardized matrix. The standardization is given by </p>
        <disp-formula id="FD29">
          <mml:math>
            <mml:mrow>
              <mml:msub>
                <mml:mi>Z</mml:mi>
                <mml:mrow>
                  <mml:mi>i</mml:mi>
                  <mml:mi>j</mml:mi>
                </mml:mrow>
              </mml:msub>
              <mml:mo>=</mml:mo>
              <mml:mfrac>
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>X</mml:mi>
                    <mml:mrow>
                      <mml:mi>i</mml:mi>
                      <mml:mi>j</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>−</mml:mo>
                  <mml:msubsup>
                    <mml:mi>μ</mml:mi>
                    <mml:mi>j</mml:mi>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mi>Ω</mml:mi>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msubsup>
                </mml:mrow>
                <mml:mrow>
                  <mml:msubsup>
                    <mml:mi>σ</mml:mi>
                    <mml:mi>j</mml:mi>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mo>(</mml:mo>
                        <mml:mi>Ω</mml:mi>
                        <mml:mo>)</mml:mo>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msubsup>
                </mml:mrow>
              </mml:mfrac>
              <mml:mo>,</mml:mo>
            </mml:mrow>
          </mml:math>
        </disp-formula>
        <p>where <inline-formula><mml:math><mml:mrow><mml:msubsup><mml:mi> μ </mml:mi><mml:mi> j </mml:mi><mml:mrow><mml:mrow><mml:mo> ( </mml:mo><mml:mi> Ω </mml:mi><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:msubsup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math><mml:mrow><mml:msubsup><mml:mi> σ </mml:mi><mml:mi> j </mml:mi><mml:mrow><mml:mrow><mml:mo> ( </mml:mo><mml:mi> Ω </mml:mi><mml:mo> ) </mml:mo></mml:mrow></mml:mrow></mml:msubsup></mml:mrow></mml:math></inline-formula> are computed from the entries for which <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> Ω </mml:mi><mml:mrow><mml:mi> i </mml:mi><mml:mi> j </mml:mi></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 1 </mml:mn></mml:mrow></mml:math></inline-formula> in variable <inline-formula><mml:math><mml:mi> j </mml:mi></mml:math></inline-formula> . In particular, the true DTC values at masked positions are excluded when calculating the DTC normalization parameters. The same observed-only parameters are then used to standardize the data for model fitting and to transform predictions back to the original scale. This transformation places different logging variables on comparable numerical scales and prevents variables with larger magnitudes from dominating the estimation of the low-rank factors.</p>
      </sec>
      <sec id="sec4dot2">
        <title>4.2. Experimental Design and Baseline Methods</title>
        <p>To evaluate the reconstruction performance under controlled conditions, artificial masking is applied to the complete DTC curves. The mask is created before calculating the standardization parameters. DTC means and standard deviations are calculated using only unmasked DTC observations; the original values at masked positions are retained solely for evaluation and are excluded both from normalization-statistic estimation and from model fitting through the observation mask matrix <inline-formula><mml:math display="inline"><mml:mi> Ω </mml:mi></mml:math></inline-formula> . Auxiliary variables remain available as specified by the experimental design.</p>
        <p><bold>Masking strategy.</bold> The masking operation is performed only on the target DTC log, while the auxiliary logs remain complete. To better approximate practical missing-data patterns, both isolated missing points and continuous missing intervals are included. The total masking ratio is 7.63% of all DTC samples. Of the 458 masked DTC samples, 295 are isolated points (64.41%) and 163 belong to continuous intervals (35.59%). For interval-length analysis, adjacent masked positions are grouped into a single continuous interval. This grouping yields 11 short intervals of 2 - 10 samples and four longer intervals of 35 - 50 samples, as shown in <xref ref-type="fig" rid="fig7">Figure 7(b)</xref>. Thus, the 2 - 10-sample group consists of adjacent masked positions that form short runs, whereas the four deliberately generated long intervals form the 35 - 50-sample group. All evaluation metrics are calculated only at the masked DTC positions, so that the results directly reflect the model’s ability to recover missing values.</p>
        <p><bold>Baseline methods.</bold> Two baseline methods are used for comparison. The first is linear interpolation, which reconstructs missing values within each well based only on local depth-wise continuity. The second is pure low-rank recovery, which uses the same matrix factorization framework as the proposed method but does not include the depth-wise smoothness correction. These two baselines are used to examine the effect of multi-log low-rank modeling and the additional contribution of depth-wise smoothness correction, respectively.</p>
        <p><bold>Parameter settings.</bold> The main parameters used in the reconstruction experiments are listed in <bold>Table 2</bold>. The rank of the low-rank factorization was set to <inline-formula><mml:math><mml:mrow><mml:mi> r </mml:mi><mml:mo> = </mml:mo><mml:mn> 3 </mml:mn></mml:mrow></mml:math></inline-formula> , which determines the dimension of the latent factor space. The regularization coefficient was fixed at <inline-formula><mml:math><mml:mrow><mml:mi> λ </mml:mi><mml:mo> = </mml:mo><mml:mn> 0.05 </mml:mn></mml:mrow></mml:math></inline-formula> to constrain the magnitude of the factor matrices and improve numerical stability. The depth-wise smoothing step size was set to <inline-formula><mml:math><mml:mrow><mml:mi> α </mml:mi><mml:mo> = </mml:mo><mml:mn> 0.05 </mml:mn></mml:mrow></mml:math></inline-formula> to control the strength of the smoothness correction. The maximum number of ALS iterations was set to <inline-formula><mml:math><mml:mrow><mml:msub><mml:mi> T </mml:mi><mml:mrow><mml:mtext> max </mml:mtext></mml:mrow></mml:msub><mml:mo> = </mml:mo><mml:mn> 5000 </mml:mn></mml:mrow></mml:math></inline-formula> . To ensure reproducibility, the random seed for initializing the factor matrices was fixed at 0.</p>
        <p>The choice of <inline-formula><mml:math><mml:mrow><mml:mi> r </mml:mi><mml:mo> = </mml:mo><mml:mn> 3 </mml:mn></mml:mrow></mml:math></inline-formula> provides a compact latent representation of the seven-variable multi-log matrix while retaining the dominant cross-log correlation structure; larger ranks introduce additional degrees of freedom and may reduce the regularizing effect of the low-rank model. The value <inline-formula><mml:math><mml:mrow><mml:mi> λ </mml:mi><mml:mo> = </mml:mo><mml:mn> 0.05 </mml:mn></mml:mrow></mml:math></inline-formula> was selected to provide moderate shrinkage of the factor matrices and to prevent numerically unstable factor magnitudes. The value <inline-formula><mml:math><mml:mrow><mml:mi> α </mml:mi><mml:mo> = </mml:mo><mml:mn> 0.05 </mml:mn></mml:mrow></mml:math></inline-formula> was selected to suppress implausible point-to-point oscillations while preserving the main depth-wise variation of the DTC curve. These parameter values were fixed before the final evaluation and were kept constant across all wells and masking patterns. They were not tuned using the true DTC values at the masked positions; those values were used only for the final evaluation.</p>
        <p><bold>Table 2.</bold> Parameter settings used in the reconstruction experiments.</p>
        <table-wrap id="tbl3">
          <label>Table 3</label>
          <table>
            <tbody>
              <tr>
                <td>Parameter</td>
                <td>Symbol</td>
                <td>Value</td>
              </tr>
              <tr>
                <td>Rank of low-rank factorization</td>
                <td>
                  <inline-formula>
                    <mml:math>
                      <mml:mi>r</mml:mi>
                    </mml:math>
                  </inline-formula>
                </td>
                <td>3</td>
              </tr>
              <tr>
                <td>Regularization coefficient</td>
                <td>
                  <inline-formula>
                    <mml:math>
                      <mml:mi>λ</mml:mi>
                    </mml:math>
                  </inline-formula>
                </td>
                <td>0.05</td>
              </tr>
              <tr>
                <td>Smoothness step size</td>
                <td>
                  <inline-formula>
                    <mml:math>
                      <mml:mi>α</mml:mi>
                    </mml:math>
                  </inline-formula>
                </td>
                <td>0.05</td>
              </tr>
              <tr>
                <td>Maximum number of ALS iterations</td>
                <td>
                  <inline-formula>
                    <mml:math>
                      <mml:mrow>
                        <mml:msub>
                          <mml:mi>T</mml:mi>
                          <mml:mrow>
                            <mml:mtext>max</mml:mtext>
                          </mml:mrow>
                        </mml:msub>
                      </mml:mrow>
                    </mml:math>
                  </inline-formula>
                </td>
                <td>5000</td>
              </tr>
              <tr>
                <td>Random seed</td>
                <td>–</td>
                <td>0</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
      </sec>
    </sec>
    <sec id="sec5">
      <title>5. Results and Discussion</title>
      <p>This section evaluates the proposed method from both quantitative and visual perspectives. The analysis includes overall reconstruction metrics, single-well curve comparisons, reconstructed-versus-true scatter plots, error distributions, well-wise error statistics, and comparisons with baseline methods. These results jointly assess the accuracy, stability, and geological plausibility of the reconstructed DTC curves.</p>
      <sec id="sec5dot1">
        <title>5.1. Overall Quantitative Performance</title>
        <p>The reconstruction performance of the proposed method is evaluated first on the masked positions of the DTC log and because the observed entries are restored to their original values during reconstruction all evaluation metrics are computed exclusively on the masked samples. The reported results thus reflect the method’s genuine missing-data recovery capability without being influenced by the observed portion of the data that remains unchanged.</p>
        <p>The quantitative evaluation of DTC reconstruction yielded the following results. The proposed method achieved an MAE of 1.810, an RMSE of 2.263, and a MAPE of 2.429%. The <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> value reached 0.848, and the correlation coefficient stood at 0.923. Taken together, these figures suggest that the reconstructed DTC values lie reasonably close to the true measurements. The MAE and RMSE remain modest, pointing to limited average errors and to the fact that larger deviations are kept in check. The low MAPE further implies that the relative reconstruction error, when viewed against the actual DTC magnitudes, is fairly small. The <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and correlation values, both relatively high.</p>
        <p>The results indicate that the proposed approach combining low-rank reconstruction with depth smoothness correction is capable of recovering missing DTC values in well-log data with reasonable effectiveness.</p>
      </sec>
      <sec id="sec5dot2">
        <title>5.2. Single-Well Curve Reconstruction Analysis</title>
        <p>A representative well is selected for closer inspection of the local reconstruction behavior, and <xref ref-type="fig" rid="fig2">Figure 2</xref> displays the juxtaposition of the true DTC curve, the masked version, and the reconstructed output.</p>
        <p><xref ref-type="fig" rid="fig2">Figure 2</xref> displays how the reconstructed DTC curve behaves in relation to the original log. Along the depth axis, the reconstruction follows the overall variation trend of the true curve with reasonable fidelity. Within the intervals that were masked, it manages to reconnect the adjacent observed segments in a manner that is relatively smooth. This outcome suggests that the method is capable of drawing upon the multi-log information available to it while still preserving a degree of continuity in the DTC response.</p>
        <p>The enlarged view further illustrates the reconstruction performance around the masked positions. In these local intervals, the reconstructed values are generally consistent with the corresponding true DTC values, indicating that the proposed method can recover the local variation pattern of the DTC curve with satisfactory accuracy. Moreover, no obvious discontinuities or unrealistic oscillations are observed in the reconstructed curve, suggesting that the depth-wise smoothness correction helps preserve local continuity across the missing intervals.</p>
        <fig id="fig2">
          <label>Figure 2</label>
          <caption>
            <p>Single-well DTC reconstruction results: overall view and local zoom-in.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId421.jpeg?20261118013139" />
        </fig>
        <p>Overall, the single-well reconstruction result provides an intuitive visualization of the proposed method’s performance. The reconstructed DTC curve not only matches the true curve closely in value, but also maintains the expected continuity along the depth direction. </p>
      </sec>
      <sec id="sec5dot3">
        <title>5.3. Scatter Plot Analysis of Reconstructed and True Values</title>
        <p>A scatter plot constructed at the masked positions, shown in <xref ref-type="fig" rid="fig3">Figure 3</xref>, further examines the agreement between the reconstructed DTC values and the true measurements.</p>
        <p><xref ref-type="fig" rid="fig3">Figure 3</xref> shows that the bulk of the scatter points lie clustered around the diagonal line. Such an alignment points to a generally favorable agreement between the reconstructed values and the true measurements, and it is consistent with the relatively high <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and Corr figures that were reported earlier in Section 5.1.</p>
        <fig id="fig3">
          <label>Figure 3</label>
          <caption>
            <p>Scatter plot of reconstructed versus true DTC values at masked positions.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId424.jpeg?20261118013140" />
        </fig>
      </sec>
      <sec id="sec5dot4">
        <title>5.4. Error Distribution Analysis</title>
        <p>To further analyze the distribution characteristics of the reconstruction errors, an error histogram is presented in <xref ref-type="fig" rid="fig4">Figure 4</xref>.</p>
        <fig id="fig4">
          <label>Figure 4</label>
          <caption>
            <p>Error distribution of DTC recovery.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId425.jpeg?20261118013140" />
        </fig>
        <p>As shown in <xref ref-type="fig" rid="fig4">Figure 4</xref>, most reconstruction errors are concentrated around zero within a relatively narrow range. This indicates that, for the majority of masked samples, the reconstructed DTC values are close to the corresponding true values, reflecting stable reconstruction behavior across the missing positions.</p>
      </sec>
      <sec id="sec5dot5">
        <title>5.5. Wellwise Stability Analysis</title>
        <p>To evaluate the cross-well stability of the proposed method, the MAE of DTC reconstruction is calculated separately for each well. The well-wise MAE results are shown in <xref ref-type="fig" rid="fig5">Figure 5</xref>.</p>
        <fig id="fig5">
          <label>Figure 5</label>
          <caption>
            <p>Well-wise MAE values of DTC reconstruction across all test wells.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId426.jpeg?20261118013141" />
        </fig>
        <p>As shown in <xref ref-type="fig" rid="fig5">Figure 5</xref>, the MAE values of different wells remain within a relatively limited range. This indicates that the proposed method does not rely on the characteristics of a single well, but maintains stable reconstruction performance across multiple wells. Such cross-well stability is important for well-log reconstruction, because logging responses may vary with geological conditions and reservoir properties.</p>
        <p>Although several wells exhibit slightly higher MAE values, the overall inter-well variation remains limited. These differences may be related to stronger local fluctuations in the DTC curves, more complex lithological changes, or the distribution of masked positions within individual wells. Nevertheless, the well-wise results show that the proposed method maintains satisfactory robustness across different wells.</p>
        <p>Overall, the well-wise MAE analysis demonstrates the cross-well applicability of the proposed reconstruction framework. </p>
      </sec>
      <sec id="sec5dot6">
        <title>5.6. Comparison with Baseline Methods</title>
        <p>Comparative experiments are conducted with two baseline methods to further evaluate the effectiveness of the proposed method: linear interpolation and pure low-rank recovery without depth-wise smoothness correction [<xref ref-type="bibr" rid="B30">30</xref>]. The quantitative comparison is summarized in <bold>Table 3</bold>, and the corresponding visual comparison is shown in <xref ref-type="fig" rid="fig6">Figure 6</xref>.</p>
        <p>As shown in <bold>Table 3</bold>, pure low-rank recovery achieves better reconstruction performance than linear interpolation. This result indicates that the low-rank factorization framework can effectively exploit the global correlation structure among multiple logging variables. Unlike linear interpolation, which relies only on local depth-wise continuity within a single log, low-rank recovery incorporates auxiliary information from other logging curves and therefore provides a more effective basis for recovering missing DTC values.</p>
        <p>The proposed method further improves the reconstruction results compared with pure low-rank recovery. Although the numerical improvements in some metrics are moderate, consistent gains are obtained in MAE, RMSE, MAPE, <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> , and Corr. This suggests that the depth-wise smoothness correction provides additional information beyond the global low-rank structure. By suppressing unrealistic local fluctuations and enhancing vertical continuity, the correction makes the reconstructed DTC curve more consistent with the expected geological behavior.</p>
        <p><bold>Table 3.</bold> Comparison of reconstruction performance among three methods for DTC curve.</p>
        <table-wrap id="tbl4">
          <label>Table 4</label>
          <table>
            <tbody>
              <tr>
                <td>Method</td>
                <td>Curve</td>
                <td>MAE</td>
                <td>RMSE</td>
                <td>MAPE (%)</td>
                <td>
                  <italic>R</italic>
                  <sup>2</sup>
                </td>
                <td>Corr</td>
              </tr>
              <tr>
                <td>Low-Rank + Smoothing</td>
                <td>DTC</td>
                <td>1.810</td>
                <td>2.263</td>
                <td>2.429</td>
                <td>0.848</td>
                <td>0.923</td>
              </tr>
              <tr>
                <td>Pure Low-Rank</td>
                <td>DTC</td>
                <td>1.833</td>
                <td>2.283</td>
                <td>2.460</td>
                <td>0.845</td>
                <td>0.922</td>
              </tr>
              <tr>
                <td>Linear Interpolation</td>
                <td>DTC</td>
                <td>2.517</td>
                <td>3.144</td>
                <td>3.395</td>
                <td>0.707</td>
                <td>0.847</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
        <fig id="fig6">
          <label>Figure 6</label>
          <caption>
            <p>Comparison of the proposed method with baseline methods for DTC reconstruction.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId429.jpeg?20261118013141" />
        </fig>
        <p>The visual comparison in <xref ref-type="fig" rid="fig6">Figure 6</xref> is consistent with the quantitative results in <bold>Table 3</bold>. Among the three methods, the proposed method achieves the lowest MAE, RMSE, and MAPE, together with the highest <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and Corr. These results show that the combination of global low-rank modeling and local depth-wise continuity correction leads to more accurate and geologically plausible DTC reconstruction.</p>
      </sec>
      <sec id="sec5dot7">
        <title>5.7. Performance Analysis under Different Missing Patterns</title>
        <p>To further examine the applicability of the proposed method under different missing-data conditions, the masked DTC samples were grouped according to the missing pattern and the length of continuous missing intervals. The corresponding RMSE comparison is presented in <xref ref-type="fig" rid="fig7">Figure 7</xref>.</p>
        <fig id="fig7">
          <label>Figure 7</label>
          <caption>
            <p>DTC reconstruction RMSE for different missing patterns and interval lengths.</p>
          </caption>
          <graphic xlink:href="https://html.scirp.org/file/7405721-rId432.jpeg?20261118013141" />
        </fig>
        <p>As shown in <xref ref-type="fig" rid="fig7">Figure 7(a)</xref>, the proposed Low-Rank + Smoothing method achieves the lowest point-level RMSE under both isolated missing points and continuous missing intervals. Compared with linear interpolation, the low-rank-based methods show a clear advantage, indicating that the correlations among multiple logging curves provide more effective information for DTC reconstruction than relying only on local depth-wise interpolation.</p>
        <p>Although the RMSE difference between Low-Rank + Smoothing and pure low-rank recovery is relatively small, the proposed method consistently yields slightly lower errors under both missing patterns. This result indicates that the low-rank structure plays the dominant role in capturing the global correlations among different logging curves and is the main source of numerical accuracy improvement. Meanwhile, the depth-wise smoothing correction further improves the local stability of the reconstructed DTC curve. For continuous missing intervals, this correction is meaningful because the reconstructed values should not only be close to the true measurements, but also remain consistent with the expected depth-wise continuity of well-log responses.</p>
        <p><xref ref-type="fig" rid="fig7">Figure 7(b)</xref> further evaluates the continuous missing intervals from a segment-level perspective. The error bars represent the standard deviation of segment-level RMSE values within each length group. The 2 - 10-sample group contains 11 runs formed by adjacent masked positions, while the 35 - 50-sample group contains the four deliberately generated long intervals. The relatively large error bar of linear interpolation in the 2 - 10 group indicates that the reconstruction difficulty varies considerably among short continuous missing intervals, suggesting that some short intervals may be located in locally fluctuating DTC sections. In the 35 - 50 group, the lower RMSE of linear interpolation does not imply that longer gaps are easier to reconstruct. Instead, it may be related to the smoother local morphology of the selected long intervals and the limited number of segments in this group. </p>
      </sec>
      <sec id="sec5dot8">
        <title>5.8. Discussion</title>
        <p>The results indicate that combining low-rank recovery with depth-wise smoothness correction is effective for reconstructing missing DTC values. The improvement over linear interpolation suggests that DTC reconstruction should not rely only on the local continuity of a single curve. By organizing multiple logs into a unified matrix, the low-rank model can exploit the correlation structure among GR, RHOB, NPHI, RT, PHIT, SW, and DTC, thereby providing additional information for estimating the masked DTC entries.</p>
        <p>The comparison with pure low-rank recovery further highlights the contribution of the depth-wise smoothness correction. Although low-rank modeling captures the global relationship among logging variables, it does not explicitly constrain the local morphology of the reconstructed curve along depth. The smoothing correction introduces the depth-continuity prior into the reconstruction process, which helps suppress unrealistic local fluctuations and improves the continuity of reconstructed DTC curves, especially within continuous missing intervals.</p>
        <p>The visual results are consistent with the quantitative evaluation. The single-well curve comparison shows that the reconstructed DTC curve follows the main trend of the true curve and reconnects masked intervals smoothly. The scatter plot and error distribution show that most reconstructed values remain close to the corresponding true values, while the well-wise MAE results indicate stable reconstruction performance across different wells. These results support the numerical accuracy and geological plausibility of the proposed framework.</p>
        <p>Some reconstruction errors are still observed, which may be related to strong local lithological variations, abrupt DTC changes, or the location and length of masked intervals. In addition, the present study focuses on DTC reconstruction under an artificial masking scheme. The applicability of the method to other logging curves, more irregular missing patterns, measurement noise, and systematic parameter selection still requires further investigation.</p>
      </sec>
    </sec>
    <sec id="sec6">
      <title>6. Conclusions</title>
      <p>This study proposed a DTC reconstruction method for incomplete well-log data by combining low-rank matrix recovery with depth-wise smoothness correction. Multiple logging variables were organized into a unified observation matrix, allowing the reconstruction model to exploit the global correlation structure among different logs. On this basis, a depth-wise smoothness correction was introduced to preserve the vertical continuity of the reconstructed DTC curve.</p>
      <p>Artificial masking experiments were conducted on complete well-log datasets to evaluate the reconstruction performance. The results were assessed using MAE, RMSE, MAPE, <inline-formula><mml:math><mml:mrow><mml:msup><mml:mi> R </mml:mi><mml:mn> 2 </mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> , and the correlation coefficient. On the masked DTC samples, the proposed method achieved satisfactory reconstruction accuracy and outperformed linear interpolation and pure low-rank recovery. These results indicate that the combination of multi-log low-rank modeling and depth-continuity correction can improve the reconstruction of missing DTC values.</p>
      <p>The masking scheme included both isolated missing points and continuous missing intervals, which allowed the proposed method to be tested under different missing-data patterns. The results show that the method can recover pointwise missing values while also maintaining reasonable curve continuity across continuous missing intervals. This suggests that the proposed framework has potential applicability to incomplete well-log datasets encountered in practical logging and reservoir evaluation.</p>
      <p>This study is limited to DTC reconstruction under artificial masking conditions. Future work may extend the proposed framework to other types of logging curves and evaluate its performance under more irregular field missing patterns, including longer and more complex depth gaps. In addition, systematic parameter selection and adaptive smoothness control deserve further investigation.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <title>References</title>
      <ref id="B1">
        <label>1.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Mukherjee, B., Sain, K. and Wu, X. (2024) Missing Log Prediction Using Machine Learning Perspectives: A Case Study from Upper Assam Basin. <italic>Earth Science Informatics</italic>, 17, 3071-3093. https://doi.org/10.1007/s12145-024-01323-5 <pub-id pub-id-type="doi">10.1007/s12145-024-01323-5</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1007/s12145-024-01323-5">https://doi.org/10.1007/s12145-024-01323-5</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Mukherjee, B.</string-name>
              <string-name>Sain, K.</string-name>
              <string-name>Wu, X.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Missing Log Prediction Using Machine Learning Perspectives: A Case Study from Upper Assam Basin</article-title>
            <source>Earth Science Informatics</source>
            <volume>17</volume>
            <pub-id pub-id-type="doi">10.1007/s12145-024-01323-5</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B2">
        <label>2.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Gama, P.H.T., Faria, J., Sena, J., Neves, F., Riffel, V.R., Perez, L., <italic>et al</italic>. (2024) Imputation in Well Log Data: A Benchmark for Machine Learning Methods. <italic>Computers &amp; Geosciences</italic>, 196, Article 105789. https://doi.org/10.1016/j.cageo.2024.105789 <pub-id pub-id-type="doi">10.1016/j.cageo.2024.105789</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.cageo.2024.105789">https://doi.org/10.1016/j.cageo.2024.105789</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Gama, P.H.T.</string-name>
              <string-name>Faria, J.</string-name>
              <string-name>Sena, J.</string-name>
              <string-name>Neves, F.</string-name>
              <string-name>Riffel, V.R.</string-name>
              <string-name>Perez, L.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Imputation in Well Log Data: A Benchmark for Machine Learning Methods</article-title>
            <source>Computers &amp; Geosciences</source>
            <volume>196</volume>
            <elocation-id>105789</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.cageo.2024.105789</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B3">
        <label>3.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Kim, M.J. and Cho, Y. (2024) Imputation of Missing Values in Well Log Data Using K-Nearest Neighbor Collaborative Filtering. <italic>Computers &amp; Geosciences</italic>, 193, Article 105712. https://doi.org/10.1016/j.cageo.2024.105712 <pub-id pub-id-type="doi">10.1016/j.cageo.2024.105712</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.cageo.2024.105712">https://doi.org/10.1016/j.cageo.2024.105712</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Kim, M.J.</string-name>
              <string-name>Cho, Y.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Imputation of Missing Values in Well Log Data Using K-Nearest Neighbor Collaborative Filtering</article-title>
            <source>Computers &amp; Geosciences</source>
            <volume>193</volume>
            <elocation-id>105712</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.cageo.2024.105712</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B4">
        <label>4.</label>
        <citation-alternatives>
          <mixed-citation publication-type="confproc">Fung, C.C., Wong, K.W., Eren, H., Charlebois, R. and Crocker, H. (1996) Modular Artificial Neural Network for Prediction of Petrophysical Properties from Well Log Data. <italic>Proceedings of IEEE Instrumentation and Measurement Technology Conference</italic>, 46, 1295-1299.</mixed-citation>
          <element-citation publication-type="confproc">
            <person-group person-group-type="author">
              <string-name>Fung, C.C.</string-name>
              <string-name>Wong, K.W.</string-name>
              <string-name>Eren, H.</string-name>
              <string-name>Charlebois, R.</string-name>
              <string-name>Crocker, H.</string-name>
            </person-group>
            <year>1996</year>
            <article-title>Modular Artificial Neural Network for Prediction of Petrophysical Properties from Well Log Data</article-title>
            <source>Proceedings of IEEE Instrumentation and Measurement Technology Conference</source>
            <volume>46</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B5">
        <label>5.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Rolon, L., Mohaghegh, S.D., Ameri, S., Gaskari, R. and McDaniel, B. (2009) Using Artificial Neural Networks to Generate Synthetic Well Logs. <italic>Journal of Natural Gas Science and Engineering</italic>, 1, 118-133. https://doi.org/10.1016/j.jngse.2009.08.003 <pub-id pub-id-type="doi">10.1016/j.jngse.2009.08.003</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.jngse.2009.08.003">https://doi.org/10.1016/j.jngse.2009.08.003</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Rolon, L.</string-name>
              <string-name>Mohaghegh, S.D.</string-name>
              <string-name>Ameri, S.</string-name>
              <string-name>Gaskari, R.</string-name>
              <string-name>McDaniel, B.</string-name>
            </person-group>
            <year>2009</year>
            <article-title>Using Artificial Neural Networks to Generate Synthetic Well Logs</article-title>
            <source>Journal of Natural Gas Science and Engineering</source>
            <volume>1</volume>
            <pub-id pub-id-type="doi">10.1016/j.jngse.2009.08.003</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B6">
        <label>6.</label>
        <citation-alternatives>
          <mixed-citation publication-type="confproc">Akinnikawe, O., Lyne, S. and Roberts, J. (2018) Synthetic Well Log Generation Using Machine Learning Techniques. <italic>Proceedings of the</italic>6 <italic>th Unconventional Resources Technology Conference</italic>, Houston, 23-25 July 2018.</mixed-citation>
          <element-citation publication-type="confproc">
            <person-group person-group-type="author">
              <string-name>Akinnikawe, O.</string-name>
              <string-name>Lyne, S.</string-name>
              <string-name>Roberts, J.</string-name>
              <string-name>Conference, H</string-name>
            </person-group>
            <year>2018</year>
            <article-title>Synthetic Well Log Generation Using Machine Learning Techniques</article-title>
            <source>Proceedings of the 6th Unconventional Resources Technology Conference</source>
            <volume>23</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B7">
        <label>7.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Zhang, D., Chen, Y. and Meng, J. (2018) Synthetic Well Logs Generation via Recurrent Neural Networks. <italic>Petroleum Exploration and Development</italic>, 45, 629-639. https://doi.org/10.1016/s1876-3804(18)30068-5 <pub-id pub-id-type="doi">10.1016/s1876-3804(18)30068-5</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/s1876-3804(18)30068-5">https://doi.org/10.1016/s1876-3804(18)30068-5</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Zhang, D.</string-name>
              <string-name>Chen, Y.</string-name>
              <string-name>Meng, J.</string-name>
            </person-group>
            <year>2018</year>
            <article-title>Synthetic Well Logs Generation via Recurrent Neural Networks</article-title>
            <source>Petroleum Exploration and Development</source>
            <volume>3804</volume>
            <issue>18</issue>
            <pub-id pub-id-type="doi">10.1016/s1876-3804(18)30068-5</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B8">
        <label>8.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Pham, N., Wu, X. and Zabihi Naeini, E. (2020) Missing Well Log Prediction Using Convolutional Long Short-Term Memory Network. <italic>Geophysics</italic>, 85, WA159-WA171. https://doi.org/10.1190/geo2019-0282.1 <pub-id pub-id-type="doi">10.1190/geo2019-0282.1</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1190/geo2019-0282.1">https://doi.org/10.1190/geo2019-0282.1</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Pham, N.</string-name>
              <string-name>Wu, X.</string-name>
              <string-name>Naeini, E.</string-name>
            </person-group>
            <year>2020</year>
            <article-title>Missing Well Log Prediction Using Convolutional Long Short-Term Memory Network</article-title>
            <source>Geophysics</source>
            <volume>85</volume>
            <pub-id pub-id-type="doi">10.1190/geo2019-0282.1</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B9">
        <label>9.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Wang, J., Cao, J., Fu, J. and Xu, H. (2022) Missing Well Logs Prediction Using Deep Learning Integrated Neural Network with the Self-Attention Mechanism. <italic>Energy</italic>, 261, Article 125270. https://doi.org/10.1016/j.energy.2022.125270 <pub-id pub-id-type="doi">10.1016/j.energy.2022.125270</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.energy.2022.125270">https://doi.org/10.1016/j.energy.2022.125270</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Wang, J.</string-name>
              <string-name>Cao, J.</string-name>
              <string-name>Fu, J.</string-name>
              <string-name>Xu, H.</string-name>
            </person-group>
            <year>2022</year>
            <article-title>Missing Well Logs Prediction Using Deep Learning Integrated Neural Network with the Self-Attention Mechanism</article-title>
            <source>Energy</source>
            <volume>261</volume>
            <elocation-id>125270</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.energy.2022.125270</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B10">
        <label>10.</label>
        <citation-alternatives>
          <mixed-citation publication-type="confproc">Kanfar, R., Shaikh, O., Yousefzadeh, M. and Mukerji, T. (2020) Real-Time Well Log Prediction from Drilling Data Using Deep Learning. <italic>International Petroleum Technology Conference</italic>, Dhahran, 13-15 January 2020. https://doi.org/10.2523/iptc-19693-ms <pub-id pub-id-type="doi">10.2523/iptc-19693-ms</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.2523/iptc-19693-ms">https://doi.org/10.2523/iptc-19693-ms</ext-link></mixed-citation>
          <element-citation publication-type="confproc">
            <person-group person-group-type="author">
              <string-name>Kanfar, R.</string-name>
              <string-name>Shaikh, O.</string-name>
              <string-name>Yousefzadeh, M.</string-name>
              <string-name>Mukerji, T.</string-name>
              <string-name>Conference, D</string-name>
            </person-group>
            <year>2020</year>
            <article-title>Real-Time Well Log Prediction from Drilling Data Using Deep Learning</article-title>
            <source>International Petroleum Technology Conference</source>
            <volume>13</volume>
            <pub-id pub-id-type="doi">10.2523/iptc-19693-ms</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B11">
        <label>11.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Wang, J., Cao, J., You, J., Cheng, M. and Zhou, P. (2021) A Method for Well Log Data Generation Based on a Spatio-Temporal Neural Network. <italic>Journal of Geophysics and Engineering</italic>, 18, 700-711. https://doi.org/10.1093/jge/gxab046 <pub-id pub-id-type="doi">10.1093/jge/gxab046</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1093/jge/gxab046">https://doi.org/10.1093/jge/gxab046</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Wang, J.</string-name>
              <string-name>Cao, J.</string-name>
              <string-name>You, J.</string-name>
              <string-name>Cheng, M.</string-name>
              <string-name>Zhou, P.</string-name>
            </person-group>
            <year>2021</year>
            <article-title>A Method for Well Log Data Generation Based on a Spatio-Temporal Neural Network</article-title>
            <source>Journal of Geophysics and Engineering</source>
            <volume>18</volume>
            <pub-id pub-id-type="doi">10.1093/jge/gxab046</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B12">
        <label>12.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Lin, L., Wei, H., Wu, T., Zhang, P., Zhong, Z. and Li, C. (2023) Missing Well-Log Reconstruction Using a Sequence Self-Attention Deep-Learning Framework. <italic>Geophysics</italic>, 88, D391-D410. https://doi.org/10.1190/geo2022-0757.1 <pub-id pub-id-type="doi">10.1190/geo2022-0757.1</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1190/geo2022-0757.1">https://doi.org/10.1190/geo2022-0757.1</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Lin, L.</string-name>
              <string-name>Wei, H.</string-name>
              <string-name>Wu, T.</string-name>
              <string-name>Zhang, P.</string-name>
              <string-name>Zhong, Z.</string-name>
              <string-name>Li, C.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Missing Well-Log Reconstruction Using a Sequence Self-Attention Deep-Learning Framework</article-title>
            <source>Geophysics</source>
            <volume>88</volume>
            <pub-id pub-id-type="doi">10.1190/geo2022-0757.1</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B13">
        <label>13.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Wang, H., Liu, X., Yang, Y. and Zhang, Y. (2024) FlexLogNet: A Flexible Deep Learning-Based Well-Log Completion Method with Heterogeneous Graph Neural Networks. <italic>Computers &amp; Geosciences</italic>, 187, Article 105611.</mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Wang, H.</string-name>
              <string-name>Liu, X.</string-name>
              <string-name>Yang, Y.</string-name>
              <string-name>Zhang, Y.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>FlexLogNet: A Flexible Deep Learning-Based Well-Log Completion Method with Heterogeneous Graph Neural Networks</article-title>
            <source>Computers &amp; Geosciences</source>
            <volume>187</volume>
            <elocation-id>105611</elocation-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B14">
        <label>14.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Cheng, C., Gao, Y., Chen, Y., Jiao, S., Jiang, Y., Yi, J., <italic>et al</italic>. (2022) Reconstruction Method of Old Well Logging Curves Based on BI-LSTM Model—Taking Feixianguan Formation in East Sichuan as an Example. <italic>Coatings</italic>, 12, Article 113. https://doi.org/10.3390/coatings12020113 <pub-id pub-id-type="doi">10.3390/coatings12020113</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/coatings12020113">https://doi.org/10.3390/coatings12020113</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Cheng, C.</string-name>
              <string-name>Gao, Y.</string-name>
              <string-name>Chen, Y.</string-name>
              <string-name>Jiao, S.</string-name>
              <string-name>Jiang, Y.</string-name>
              <string-name>Yi, J.</string-name>
            </person-group>
            <year>2022</year>
            <article-title>Reconstruction Method of Old Well Logging Curves Based on BI-LSTM Model—Taking Feixianguan Formation in East Sichuan as an Example</article-title>
            <source>Coatings</source>
            <volume>12</volume>
            <elocation-id>113</elocation-id>
            <pub-id pub-id-type="doi">10.3390/coatings12020113</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B15">
        <label>15.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Feng, R., Grana, D. and Balling, N. (2021) Imputation of Missing Well Log Data by Random Forest and Its Uncertainty Analysis. <italic>Computers &amp; Geosciences</italic>, 152, Article 104763. https://doi.org/10.1016/j.cageo.2021.104763 <pub-id pub-id-type="doi">10.1016/j.cageo.2021.104763</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.cageo.2021.104763">https://doi.org/10.1016/j.cageo.2021.104763</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Feng, R.</string-name>
              <string-name>Grana, D.</string-name>
              <string-name>Balling, N.</string-name>
            </person-group>
            <year>2021</year>
            <article-title>Imputation of Missing Well Log Data by Random Forest and Its Uncertainty Analysis</article-title>
            <source>Computers &amp; Geosciences</source>
            <volume>152</volume>
            <elocation-id>104763</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.cageo.2021.104763</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B16">
        <label>16.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Tatsipie, N.R.K. and Sheng, J.J. (2021) Generating Pseudo Well Logs for a Part of the Upper Bakken Using Recurrent Neural Networks. <italic>Journal of Petroleum Science and Engineering</italic>, 200, Article 108253. https://doi.org/10.1016/j.petrol.2020.108253 <pub-id pub-id-type="doi">10.1016/j.petrol.2020.108253</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1016/j.petrol.2020.108253">https://doi.org/10.1016/j.petrol.2020.108253</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Tatsipie, N.R.K.</string-name>
              <string-name>Sheng, J.J.</string-name>
            </person-group>
            <year>2021</year>
            <article-title>Generating Pseudo Well Logs for a Part of the Upper Bakken Using Recurrent Neural Networks</article-title>
            <source>Journal of Petroleum Science and Engineering</source>
            <volume>200</volume>
            <elocation-id>108253</elocation-id>
            <pub-id pub-id-type="doi">10.1016/j.petrol.2020.108253</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B17">
        <label>17.</label>
        <citation-alternatives>
          <mixed-citation publication-type="confproc">Gunasekar, S., Yamada, M., Yin, D. and Chang, Y. (2015) Consistent Collective Matrix Completion under Joint Low Rank Structure. <italic>Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics</italic>, 38, 306-314.</mixed-citation>
          <element-citation publication-type="confproc">
            <person-group person-group-type="author">
              <string-name>Gunasekar, S.</string-name>
              <string-name>Yamada, M.</string-name>
              <string-name>Yin, D.</string-name>
              <string-name>Chang, Y.</string-name>
            </person-group>
            <year>2015</year>
            <article-title>Consistent Collective Matrix Completion under Joint Low Rank Structure</article-title>
            <source>Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics</source>
            <volume>38</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B18">
        <label>18.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Wang, Z., Lai, M., Lu, Z., Fan, W., Davulcu, H. and Ye, J. (2015) Orthogonal Rank-One Matrix Pursuit for Low Rank Matrix Completion. <italic>SIAM Journal on Scientific Computing</italic>, 37, A488-A514. https://doi.org/10.1137/130934271 <pub-id pub-id-type="doi">10.1137/130934271</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1137/130934271">https://doi.org/10.1137/130934271</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Wang, Z.</string-name>
              <string-name>Lai, M.</string-name>
              <string-name>Lu, Z.</string-name>
              <string-name>Fan, W.</string-name>
              <string-name>Davulcu, H.</string-name>
              <string-name>Ye, J.</string-name>
            </person-group>
            <year>2015</year>
            <article-title>Orthogonal Rank-One Matrix Pursuit for Low Rank Matrix Completion</article-title>
            <source>SIAM Journal on Scientific Computing</source>
            <volume>37</volume>
            <pub-id pub-id-type="doi">10.1137/130934271</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B19">
        <label>19.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Wang, Y., Xia, K., Niu, W. and Ahamd, A. (2017) A Denoising Method by Low-Rank Matrix Recovery and Its Application in Oil Well Logging. <italic>Geophysical Prospecting</italic><italic>for Petroleum</italic>, 56, 644-650.</mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Wang, Y.</string-name>
              <string-name>Xia, K.</string-name>
              <string-name>Niu, W.</string-name>
              <string-name>Ahamd, A.</string-name>
            </person-group>
            <year>2017</year>
            <article-title>A Denoising Method by Low-Rank Matrix Recovery and Its Application in Oil Well Logging</article-title>
            <source>Geophysical Prospecting for Petroleum</source>
            <volume>56</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B20">
        <label>20.</label>
        <citation-alternatives>
          <mixed-citation publication-type="confproc">Acharya, S. and Fabian, K. (2023) Dynamic Depth Alignment of Well-Bore Measurements Using Machine Learning. 84 <italic>th EAGE Annual Conference &amp; Exhibition</italic>, 2023, 1-5. https://doi.org/10.3997/2214-4609.2023101047 <pub-id pub-id-type="doi">10.3997/2214-4609.2023101047</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3997/2214-4609.2023101047">https://doi.org/10.3997/2214-4609.2023101047</ext-link></mixed-citation>
          <element-citation publication-type="confproc">
            <person-group person-group-type="author">
              <string-name>Acharya, S.</string-name>
              <string-name>Fabian, K.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Dynamic Depth Alignment of Well-Bore Measurements Using Machine Learning</article-title>
            <source>84th EAGE Annual Conference &amp; Exhibition</source>
            <volume>2023</volume>
            <pub-id pub-id-type="doi">10.3997/2214-4609.2023101047</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B21">
        <label>21.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Westeng, K., Aursand, P., Viset, F. and Van Crombrugge, Y. (2026) Dynamic Depth Alignment of Well Logs: A Continuous Optimization Framework for Enhanced Petrophysical and Rock Physics Interpretation. <italic>Petrophysics</italic>— <italic>The SPWLA Journal of Formation Evaluation and Reservoir Description</italic>, 67, 54-67. https://doi.org/10.30632/pjv67n1-2026a4 <pub-id pub-id-type="doi">10.30632/pjv67n1-2026a4</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.30632/pjv67n1-2026a4">https://doi.org/10.30632/pjv67n1-2026a4</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Westeng, K.</string-name>
              <string-name>Aursand, P.</string-name>
              <string-name>Viset, F.</string-name>
              <string-name>Crombrugge, Y.</string-name>
            </person-group>
            <year>2026</year>
            <article-title>Dynamic Depth Alignment of Well Logs: A Continuous Optimization Framework for Enhanced Petrophysical and Rock Physics Interpretation</article-title>
            <source>Petrophysics—The SPWLA Journal of Formation Evaluation and Reservoir Description</source>
            <volume>67</volume>
            <pub-id pub-id-type="doi">10.30632/pjv67n1-2026a4</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B22">
        <label>22.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Pilkington, M. and Todoeschuck, J.P. (1991) Naturally Smooth Inversions with a Priori Information from Well Logs. <italic>Geophysics</italic>, 56, 1811-1818. https://doi.org/10.1190/1.1442993 <pub-id pub-id-type="doi">10.1190/1.1442993</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1190/1.1442993">https://doi.org/10.1190/1.1442993</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Pilkington, M.</string-name>
              <string-name>Todoeschuck, J.P.</string-name>
            </person-group>
            <year>1991</year>
            <article-title>Naturally Smooth Inversions with a Priori Information from Well Logs</article-title>
            <source>Geophysics</source>
            <volume>56</volume>
            <pub-id pub-id-type="doi">10.1190/1.1442993</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B23">
        <label>23.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Wu, X. and Caumon, G. (2017) Simultaneous Multiple Well-Seismic Ties Using Flattened Synthetic and Real Seismograms. <italic>Geophysics</italic>, 82, IM13-IM20. https://doi.org/10.1190/geo2016-0295.1 <pub-id pub-id-type="doi">10.1190/geo2016-0295.1</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1190/geo2016-0295.1">https://doi.org/10.1190/geo2016-0295.1</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Wu, X.</string-name>
              <string-name>Caumon, G.</string-name>
            </person-group>
            <year>2017</year>
            <article-title>Simultaneous Multiple Well-Seismic Ties Using Flattened Synthetic and Real Seismograms</article-title>
            <source>Geophysics</source>
            <volume>82</volume>
            <pub-id pub-id-type="doi">10.1190/geo2016-0295.1</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B24">
        <label>24.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Cai, J., Candès, E.J. and Shen, Z. (2010) A Singular Value Thresholding Algorithm for Matrix Completion. <italic>SIAM Journal on Optimization</italic>, 20, 1956-1982. https://doi.org/10.1137/080738970 <pub-id pub-id-type="doi">10.1137/080738970</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1137/080738970">https://doi.org/10.1137/080738970</ext-link></mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Cai, J.</string-name>
              <string-name>Shen, Z.</string-name>
            </person-group>
            <year>2010</year>
            <article-title>A Singular Value Thresholding Algorithm for Matrix Completion</article-title>
            <source>SIAM Journal on Optimization</source>
            <volume>20</volume>
            <pub-id pub-id-type="doi">10.1137/080738970</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B25">
        <label>25.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Shi, L. and Zou, C. (2020) Noisy Low‐Rank Matrix Completion under General Bases. <italic>Stat</italic>, 9, e303. https://doi.org/10.1002/sta4.303 <pub-id pub-id-type="doi">10.1002/sta4.303</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1002/sta4.303">https://doi.org/10.1002/sta4.303</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Shi, L.</string-name>
              <string-name>Zou, C.</string-name>
            </person-group>
            <year>2020</year>
            <article-title>Noisy Low‐Rank Matrix Completion under General Bases</article-title>
            <source>Stat</source>
            <volume>9</volume>
            <pub-id pub-id-type="doi">10.1002/sta4.303</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B26">
        <label>26.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Yang, L., Wang, S., Chen, X., Chen, W., Saad, O.M. and Chen, Y. (2023) Deep-Learning Missing Well-Log Prediction via Long Short-Term Memory Network with Attention-Period Mechanism. <italic>Geophysics</italic>, 88, D31-D48. https://doi.org/10.1190/geo2020-0749.1 <pub-id pub-id-type="doi">10.1190/geo2020-0749.1</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1190/geo2020-0749.1">https://doi.org/10.1190/geo2020-0749.1</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Yang, L.</string-name>
              <string-name>Wang, S.</string-name>
              <string-name>Chen, X.</string-name>
              <string-name>Chen, W.</string-name>
              <string-name>Saad, O.M.</string-name>
              <string-name>Chen, Y.</string-name>
            </person-group>
            <year>2023</year>
            <article-title>Deep-Learning Missing Well-Log Prediction via Long Short-Term Memory Network with Attention-Period Mechanism</article-title>
            <source>Geophysics</source>
            <volume>88</volume>
            <pub-id pub-id-type="doi">10.1190/geo2020-0749.1</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B27">
        <label>27.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Keshavan, R.H., Montanari, A. and Oh, S. (2010) Matrix Completion from a Few Entries. <italic>IEEE Transactions on Information Theory</italic>, 56, 2980-2998. https://doi.org/10.1109/tit.2010.2046205 <pub-id pub-id-type="doi">10.1109/tit.2010.2046205</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1109/tit.2010.2046205">https://doi.org/10.1109/tit.2010.2046205</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Keshavan, R.H.</string-name>
              <string-name>Montanari, A.</string-name>
              <string-name>Oh, S.</string-name>
            </person-group>
            <year>2010</year>
            <article-title>Matrix Completion from a Few Entries</article-title>
            <source>IEEE Transactions on Information Theory</source>
            <volume>56</volume>
            <pub-id pub-id-type="doi">10.1109/tit.2010.2046205</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B28">
        <label>28.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Zhao, L., Babu, P. and Palomar, D.P. (2016) Efficient Algorithms on Robust Low-Rank Matrix Completion against Outliers. <italic>IEEE Transactions on Signal Processing</italic>, 64, 4767-4780. https://doi.org/10.1109/tsp.2016.2572049 <pub-id pub-id-type="doi">10.1109/tsp.2016.2572049</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.1109/tsp.2016.2572049">https://doi.org/10.1109/tsp.2016.2572049</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Zhao, L.</string-name>
              <string-name>Babu, P.</string-name>
              <string-name>Palomar, D.P.</string-name>
            </person-group>
            <year>2016</year>
            <article-title>Efficient Algorithms on Robust Low-Rank Matrix Completion against Outliers</article-title>
            <source>IEEE Transactions on Signal Processing</source>
            <volume>64</volume>
            <pub-id pub-id-type="doi">10.1109/tsp.2016.2572049</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B29">
        <label>29.</label>
        <citation-alternatives>
          <mixed-citation publication-type="journal">Hastie, T., Mazumder, R., Lee, J.D. and Zadeh, R. (2015) Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares. <italic>Journal of Machine Learning Re-search</italic>, 16, 3367-3402.</mixed-citation>
          <element-citation publication-type="journal">
            <person-group person-group-type="author">
              <string-name>Hastie, T.</string-name>
              <string-name>Mazumder, R.</string-name>
              <string-name>Lee, J.D.</string-name>
              <string-name>Zadeh, R.</string-name>
            </person-group>
            <year>2015</year>
            <article-title>Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares</article-title>
            <source>Journal of Machine Learning Re-search</source>
            <volume>16</volume>
          </element-citation>
        </citation-alternatives>
      </ref>
      <ref id="B30">
        <label>30.</label>
        <citation-alternatives>
          <mixed-citation publication-type="other">Wang, G., Teng, H., Qiao, L., Yu, H., Cui, Y. and Xiao, K. (2024) Well Logging Reconstruction Based on a Temporal Convolutional Network and Bidirectional Gated Recurrent Unit Network with Attention Mechanism Optimized by Improved Sand Cat Swarm Optimization. <italic>Energies</italic>, 17, Article 2710. https://doi.org/10.3390/en17112710 <pub-id pub-id-type="doi">10.3390/en17112710</pub-id><ext-link ext-link-type="uri" xlink:href="https://doi.org/10.3390/en17112710">https://doi.org/10.3390/en17112710</ext-link></mixed-citation>
          <element-citation publication-type="other">
            <person-group person-group-type="author">
              <string-name>Wang, G.</string-name>
              <string-name>Teng, H.</string-name>
              <string-name>Qiao, L.</string-name>
              <string-name>Yu, H.</string-name>
              <string-name>Cui, Y.</string-name>
              <string-name>Xiao, K.</string-name>
            </person-group>
            <year>2024</year>
            <article-title>Well Logging Reconstruction Based on a Temporal Convolutional Network and Bidirectional Gated Recurrent Unit Network with Attention Mechanism Optimized by Improved Sand Cat Swarm Optimization</article-title>
            <source>Energies</source>
            <volume>17</volume>
            <elocation-id>2710</elocation-id>
            <pub-id pub-id-type="doi">10.3390/en17112710</pub-id>
          </element-citation>
        </citation-alternatives>
      </ref>
    </ref-list>
  </back>
</article>