Article citationsMore>>

Jia, X. and Guo, B. (2022) Non-Negative Matrix Factorization Based on Smoothing and Sparse Constraints for Hyperspectral Unmixing. Sensors, 22, Article 5417.
https://doi.org/10.3390/s22145417

has been cited by the following article:

• TITLE: Optimal Predictive Modeling of Nonlinear Transformations: Innovative Applied Mathematics in an Artificial Intelligence System

  AUTHORS: Jean Pacifique Nkurunziza, Fulgence Nahayo, Ignat Anca, Thierry Nsabimana

  KEYWORDS: Optimal Predictive Modeling, Nonlinear Transformation, Artificial Intelligence Systems, Parameter Optimization, KL-ONMF Model, Clustering, Cross-Validation

  JOURNAL NAME: Open Journal of Applied Sciences, Vol.15 No.10, October 15, 2025

  ABSTRACT: In this paper, an Optimal Predictive Modeling of Nonlinear Transformations “OPMNT” method has been developed while using Orthogonal Nonnegative Matrix Factorization “ONMF” with the Kullback-Leibler “KL” divergence. Innovative algorithm includes a Term Frequency-Inverse Document Frequency score smoothing technique and incorporates parameters optimized through cross-validation using the Nelder-Mead method. This technique has numerous applications in Artificial Intelligence Systems (AIS), including extracting nonlinear features for document classification. To make our technique powerful, we apply nonlinear transformations such as logarithms, square roots, and hyperbolic tangents. Our results demonstrate that the OPMNT-KL-ONMF method yields significantly higher accuracy on average compared to untransformed datasets, underscoring the critical importance of selecting appropriate transformation functions to enhance the classification capabilities of the KL-ONMF model. Future work will involve integrating these transformation functions into a neural network framework to explore new techniques for improving performance.