The method of science calls for the understanding of selected aspects of behaviour of a considered system, given available measurements and other relevant information. The measurements may be of the variable while the parameters that define the selected system behaviour may be, or the selected system behaviour can itself be an unknown and sought function of the known input variable vector , so that. In either case, we relate the measurements with the model of the system behaviour as in the equation or where the function is unknown. Alternatively, in either case the scientist aims to solve an inverse problem in which the operator, when operated upon the data, yields the unknown(s).

One problem that then immediately arises is the learning of the unknown function. Indeed is often unknown though such is not the norm―for example in applications in which the data is generated by a known projection of the model function onto the space of the measurables, is identified as this known projection. Thus, image inversion is an example of an inverse problem in which the data are a known function of the unknown model function or model parameter vector [1-5, among others]. On the other hand, there can arise a plethora of other situations in science in which a functional relationship between the measurable and unknown (or) is appreciated but the exact form of this functional relationship is not known [6-12, to cite a few].

This situation allows for a (personal) classification of inverse problems such that

• in inverse problems of Type I, is known where or,

• in inverse problems of Type II, is unknown.

While inverse problems of Type I can be rendered difficult owing to these being ill-posed and/or ill- conditioned as well as in the quantification of the uncertainties in the estimation of the unknown(s), inverse problems of Type II appear to be entirely intractable in the current formulation of (or), where the aim is the learning of the unknown (or), given the data. In fact, conventionally, this very general scientific problem would not even be treated as an inverse problem but rather as a modelling exercise specific to the relevant scientific discipline. From the point of view of inverse problems, these entails another layer of learning, namely, the learning of from the data―to be precise, from training data [13-15]. Here by training data we mean data that comprises values of at chosen values of (or at chosen). These chosen (and therefore known) values of (or) are referred to as the design points, so that values of generated for the whole design set comprise the training data.Having trained the model for using such training data, we then implement this learnt model on the available mea- surements―or test data―to learn that value of (or) at which the measurements are realised.

It is in principle possible to generate a training data set from surveys (as in selected social science applications) or generate synthetic training data sets using simulation models of the system [16-18]. However, often the Physics of the situation is such that is rendered characteristic of the system at hand (as in complex physical and biological systems). Consequently, a simulation model of the considered system is only an approximation of the true underlying Physics and therefore risky in general; after all, the basic motivation behind the learning of the unknown (or) is to learn the underlying system Physics, and pivoting such learning on a simulation model that is of unquantifiable crudeness, may not be useful.

Thus, in such cases, we need to develop an alternative way of learning or if possible, learn the un- known (or) given the available measurements without needing to know. It may appear that such is possible in the Bayesian approach in which we only need to write the posterior probability density of the unknown (or), given the data. An added advantage of using the Bayesian framework is that extra information is brought into the model via the priors, thus reducing the quantity of data required to achieve inference of a given quality. Importantly, in this approach one can readily achieve estimation of uncertainties in the relevant parameters, as distinguished from point estimates of the same. In this paper, we present the Bayesian learning of the unknown model parameters given the measurements but no training data, as no training data set is available. The presented methodology is inspired by state space modelling techniques and is elucidated using an application to astronomical data.

The advantages of the Bayesian framework notwithstanding, in systems in which training data is unavailable, fact remains that cannot be learnt. This implies that if learning of the unknown (or) is attempted by modelling as a realisation from a stochastic process (such as a Gaussian Process (GP) or Ito Process or t-process, etc.), then the correlation structure that underlies this process is not known. However, in this learning approach, the posterior probability of the unknowns given the data invokes such a correlation structure. Only by using training data can we learn the covariance of the process that is sampled from, leading to our formulation of the posterior of the unknowns, given the measured data as well as the training data. To take the example of modelling using a high-dimensional GP, it might be possible of course to impose the form of the covariance by hand; for example, when it is safe to assume that is continuous, we could choose a stationary covariance function [19], such as the popular square exponential covariance or the Matern class of covariance functions [20], though parameters of such a covariance (correlation length, smoothness parameter) being unknown, ad hoc values of these will then need to be imposed. In the presence of training data, the smoothness parameters can be learnt from the data.

For systems in which the continuous assumption is misplaced, choosing an appropriate covariance function and learning the relevant parameters from the measured data, in absence of training data, becomes even trickier. An example of this situation can arise in fact in an inverse problem of Type I―the unknown physical density of the system is projected onto the space of observables such that inversion of the available (noisy) image data will allow for the estimation of the unknown density, where the projection operator is known. Such a density function in real systems can often have disjoint support in its domain and can also be typically characterised by sharp density contrasts as in material density function of real-life material samples [21]. Then, if we were to model this discontinuous and multimodal density function as a realisation from a GP, the covariance function of such a process will need to be non-stationary. It is possible to render a density function sampled from such a GP to be differently differentiable at different points, using for example prescriptions advanced in the literature [22], but in lieu of training data it is not possible to parametrise covariance kernels to ensure the representative discontinuity and multimodality of the sampled (density) functions. Thus, the absence of training data leads to the inability to learn the correlation structure of the density function given the measured image data.

A way out this problem could be to make an attempt to construct a training data set by learning values of the unknown system behaviour function at those points in the domain of the density, at which measured data are available; effectively, we then have a set of data points, each generated at a learnt value of the function, i.e. this set comprises a training data. In this data set there are measurement uncertainties as well as uncertainty of estimation on each of the learnt values of the system function. Of course, learning the value of the function at identified points within the domain of the system function, is in itself a difficult task. Thus, in this paradigm, the domain of the unknown system function is discretised according to the set of values of, , at which the measurements are available. In other words, the discretisation of is dictated by the data distribution. Over each -bin, the function is held a constant such that for in the i-th bin, the function takes the value,; then we define and try to learn this vector, given the data. Unless otherwise motivated, in general applications, the probability distribution of is not imposed by hand. In the Bayesian framework this exercise translates to the computing of the joint posterior probability density of distribution-free parameters given the data, where the correlation between and is not invoked,. Of course, framed this way, we can only estimate the value of the sought function at identified values of―unless interpolation is used―but once the training data, thus constructed, is subsequently implemented in the modelling of with a GP of appropriate dimensionality, statistical prediction at any value of may be possible.

Above, we dealt schematically with the difficult case of lack of training data. However, even when a training data set is available, learning using such data can be hard. In principle, can be learnt using splines or wavelets. However, a fundamental shortcoming of this method is that splines and wavelets can fail to capture the correlation amongst the component functions of a high-dimensional. Also, the numerical difficulty of the very task of learning using this technique, and particularly of inverting the learnt, only increases with dimensionality. Thus it is an improvement to model such a with a high-dimensional GP. A high-dimensional can arise in a real-life inverse problem if the observed data is high-dimensional, eg. the data is matrix-variate [23].

Measurement uncertainties or measurement noise is almost unavoidable in practical applications and therefore, any attempt at an inference on the unknown model parameter vector (or the unknown model function) should be capable of folding in such noise in the data. In addition to this, there could be other worries stemming from inadequacies of the available measurements―the data could be “too small” to allow for any meaningful inference on the unknown(s) or “too big” to allow for processing within practical time frames; here the qualification of the size of the data is determined by the intended application as well as the constraints on the available computational resources. However, a general statement that is relevant here is the fact that in the Bayesian paradigm, less data is usually required than in the frequentists’ approach, as motivated above. Lastly, data could also be missing; in particular, in this paper we discuss a case in which the measurable lives in a space where is the state space of the system at hand.

The paper is constructed as follows. In Section 2, we briefly discuss the outline of state space modelling. In the following Section 3, our new state space modelling based methodology is delineated; in particular, we explore alternatives to the suggested method in Subsection 3.1. The astrophysical background to the application using which our methodology is elucidated, is motivated in Section 4 while the details of the modelling are presented in Section 5. We present details of our inference in Section 6 and applications to synthetic and real data are considered in Section 7 and Section 8 respectively. We round up the paper with some discussions about the ramifications of our results in Section 9.

Understanding the evolution of the probability density function of the state space of a dynamical system, given the available data, is of broad interest to practitioners across disciplines. Estimation of the parameters that affect such evolution can be performed within the framework of state space models or SSMs [24-27]. Basically, an SSM comprises an observation structure and an evolution structure. Assuming the observations to be con- ditionally independent, the marginal distribution of any observation is dependent on a known or unknown stationary model parameter, at a given value of the state space parameter at the current time. Modelling of errors of such observations within the SSM framework is of interest in different disciplines [28,29].

The evolution of the state space parameter is on the other hand given by another set of equations, in which the uncertainty of the evolved value of the parameter is acknowledged. A state space representation of complex systems will in general have to be designed to capacitate high-dimensional inference in which both the evolutionary as well as observation equations are in general non-linear and parameters and uncertainties are non-Gaussian.

In this paper we present a new methodology that offers a state space representation in a situation when data is collected at only one time point and the unknown state space parameter in this treatment is replaced by the discretised version of the multivariate probability density function (pdf) of the state space variable. The focus is on the learning of the static unknown model parameter vector rather than on prediction of the state space parameter at a time point different to when the observations are made. In fact, the sought model parameter vector is treated as embedded within the definition of the pdf of the state space variable. In particular, the method that we present here pertains to a partially observed state space, i.e. the observations comprise measurements on only some―but not all―of the components of the state space vector. Thus in this paradigm, probability of the observations conditional on the state space parameters reduces to the probability that the observed state space data have been sampled from the pdf of the full state space variable vector, marginalised over the unobserved components. Here this pdf includes the sought static model parameter vector in its definition. In addition to addressing missing data, the presented methodology is developed to acknowledge the measurement errors that may be non-Gaussian.

The presented method is applied to real and synthetic astronomical data with the aim of drawing inference on the distribution of the gravitational mass of all matter in a real and simulated galaxy, respectively. This gravitational mass density is projected to be useful in estimating the distribution of dark matter in the galactic system.

Here we aim to learn the unknown model parameter vector given the data, where data comprises measurements of some (h) components of the d-dimensional state space parameter vector; thus,. Here. In fact, the data set is where the i-th observation is the vector. Let the state space be so that. Let the observable vector be. Let, i.e. the probability density function of the state parameter vector is, where the distribution is parametrised by the parameter.

In light of this, we suggest that. Then had the observations lived in the state space, we could have advanced the likelihood function in terms of. However, here we deal with missing data that we know lives in the sub-space within. Therefore, the data must be sampled from the density that is obtained by marginalising the pdf over. In other words, the pdf is projected onto the space of the observables, i.e. onto; the result is the projected or marginalised density of the observables. Then under the assumption of the observed vectors being conditionally iid, the likelihood function is

where

While the likelihood is thus defined, what this definition still does not include in it is the sought model parameter vector. In this treatment, we invoke a secondary equation that allows for the model parameter vector to be embedded into the definition of the likelihood. This can be accomplished by eliciting application specific details but in general, we suggest and construct the general model for the state space pdf to be

where is a t-dimensional vector function of a vector.

Given this rephrasing of the state space pdf, the projected density that the i-th measurement is sampled from, is re-written as

so that plugging this in the RHS of Equation (1), the likelihood is

However, it is appreciated that the pdf of the state space vector may not be known, i.e. is unknown. This motivates us to attempt to learn the state space pdf from the data, simultaneously with. We consider the situation that training data is unavailable where training data would comprise a set of values of generated at chosen values of. However, since the very functional relationship (in the notation motivated above) between and is not known, it is not possible to generate values of at a chosen value of, unless of course, an approximation of unquantifiable crudeness for this functional relationship is invoked. Here we attempt to improve upon the prospect of imposing an ad hoc model of. Then in this paradigm, we discretise the function.

This is done by placing the relevant ranges of the vectors and on a grid of chosen cell size. Thus, for and being discretised into and j-dimensional vectors respectively, the discretised version of is then represented as the -dimensional vector such that the p-th component of this vector is the value of in the p-th “-grid cell”. Here, such a grid-cell is the p-th of the ones that the domain of is discretised into,.

Given this discretisation of, the RHS of Equation (4) is reduced to a sum of integrals over the unobserved variable in each of the grid-cells. In other words,

where is the value that the vector of the unobserved variables takes up in the p-th -grid-cell. The integral on the RHS of Equation (6) represents the volume that the p-th -grid-cell occupies in the space of the unobserved variable vector. The value of in the p-th -grid-cell is dependent in general on for a given data vector; hence the notation.

In other words, to compute the integral for each p (on the RHS of Equation (6)) we need to identify the bounds on the value of each component of imposed by the edges of the p-th grid-cell. This effectively calls for identification of the mapping between the space of and, and the space of the unobserved variables. Now the observation. Then, where. Indeed, this mapping will be understood using the physics of the system at hand. We will address this in detail in the context of the application that is considered in the paper.

The likelihood function is then again rephrased as

using Equation (6).

However, the observed data is likely to be noisy too. To incorporate the errors of measurement, the likelihood is refined by convolving with the density of the error in the value of the observed vector, where the error distribution is assumed known. Let the density of the error distribution be where are the known parameters. Then the likelihood is finally advanced as

In a Bayesian framework, inference is pursued thereafter by selecting priors for the unknowns and, and then using the selected priors in conjunction with the likelihood defined in Equation 8, in Bayes rule to give

the posterior of the unknowns given the data, i.e.. In the context of the application at hand,

we will discuss all this and in particular, advance the data-driven choice of the details of the discretisation of the function. Posterior samples could be generated using a suitable version of Metropolis-Hastings and implemented to compute the 95% HPD credible regions on the learnt parameter values.

Alternative methods

We ask ourselves the question about alternative treatment of the data that could result in the estimation of the unknown model parameter vector. Let the sought model parameter be s-dimensional while the observable is an h-dimensional vector valued variable and there are number of measurements of this variable available. Then the pursuit of can lead us to express the data as a function of the model parameter vector, i.e. write, where is an unknown, h-dimensional vector valued function of an s-dimensional vector. In order to learn, we will need to first learn from the data, as was motivated in the introductory section.

As we saw in that section, the learning of this high-dimensional function from the data and its inversion are best tackled by modelling the unknown high-dimensional function with a Gaussian Process. [23] present a generic Bayesian method that performs the learning and inversion of a high-dimensional function given matrix-variate data within a supervised learning paradigm; the (chosen) stationary covariance function implemented in this work is learnt using training data and is subsequently used in the computation of the posterior probability of the unknown model parameter vector given the measured or test data, as well as the training data. In the absence of available training data, such an implementation is not possible, i.e. such a method is not viable in the unsupervised learning paradigm. In the application we discuss below, training data is not available and therefore, the modelling of the functional relation between data and, using Gaussian Processes appears to not be possible. This shortcoming can however be addressed if simulations of the system at hand can be undertaken to yield data at chosen values of; however, the very physical mechanism that connects with the data may be unknown (as in the considered application) and therefore, such a simulation model is missing. Alternatively, if independently learnt, learnt with an independent data set, is available, the same can be used as training data to learn given another data set. On such instances, the Gaussian Process approach is possible but in lieu of such training data becoming available, the learning of given the matrix-valued data can be performed in the method presented above. On the other hand, a distinct advantage of the method presented below is that it allows for the learning of the state space density in addition to the unknown model parameter vector.

If the suggestion is to learn the unknown system function as itself a realisation of a GP, the question that then needs to be addressed is how to parametrise the covariance structure of GP in situations in which the data results from measurements of the variable that shares an unknown functional relation with. In other words, in such situations, the unknown system function has to be linked with the available data via a functional relation, which however is unknown, as motivated above; we are then back to the discussion in the previous paragraph.

Unravelling the nature of Dark Matter and Dark Energy is one of the major challenges of today’s science. While such is pursued, the gathering of empirical evidence for/against Dark Matter (DM) in individual real-life observed astronomical systems is a related interesting exercise.

The fundamental problem in the quantification of dark matter in these systems is that direct observational evidence of DM remains elusive. In light of this, the quantification is pursued using information obtained from measurable physical manifestations of the gravitational field of all matter in an astronomical system, i.e. dark as well as self-luminous matter. Indeed, such measurements are difficult and physical properties that manifest the gravitational effect of the total gravitational field of the system would include the density of X-rays emitted by the hot gas in the system at a measured temperature [30], velocities of individual particles that live in the system and play in its gravitational field [31-35] and the deviation in the path of a ray of light brought about by the gravitational field of the system acting as a gravitational lens [36].

The extraction of the density of DM from the learnt total gravitational mass density of all matter in the system, is performed by subtracting from the latter, the gravitational mass density of the self-luminous matter. The density of such luminous matter is typically modelled astronomically using measurements of the light that is observed from the system. A reliable functional relationship between the total gravitational mass density and such photometric measurements is not motivated by any physical theories though the literature includes such a relationship as obtained from a pattern recognition study performed with a chosen class of galaxies [37].

In this work, we focus our attention to the learning of the total gravitational mass density in galaxies, the images of which resemble ellipses―as distinguished from disc-shaped galaxies for which the sought density is more easily learnt using measurement of rotational speed of resident particles. By a galactic “particle” we refer to resolved galactic objects such as stars. There could also be additional types of particles, such as planetary nebulae (PNe) which are an end state of certain kinds of stars; these bear signature marks in the emitted spectral data. Other examples of galactic particles could include old clusters of stars, referred to as globular clusters (GCs).

Data

As defined above, the space of all states that a dynamical system achieves is referred to as the system’s state space. Now, the state that a galaxy is in, is given by the location and velocity coordinates of all particles in the system. Here, the location coordinate is as is the velocity coordinate vector. Thus, in our treatment of the galaxy at hand, is the space of the particle location and velocity vector i.e. the space of the vector. We model the galactic particles to be playing in the average (gravitational) force field that is given rise to by all the particles in this system. Under the influence of this mean field, we assume the system to have relaxed to a stationary state so that there is no time dependence in the distribution of the vector

, where the 3-dimensional vector and. Then the pdf of the variable is, where is a parameter vector.

Our aim is to learn the density function of gravitational mass of all matter in the galaxy, given the data

, where. The physical interpretation of these observables is that is the

component of the velocity of a galactic particle that is aligned along the line-of-sight that joins the particle and the observer, i.e. we can measure how quickly the particle is coming towards the observer or receding away but cannot measure any of the other components of. Similarly, we know the components and of the location of a galactic particle in the galactic image but cannot observe how far orthogonal to the image plane the

particle is, i.e. is unobservable. Thus but with. It merits mention that in the available data, values of and appear in the form of. Then the data.

Here is typically of the order of 10². While for more distant galaxies, is lower, recent advancements is astronomical instrumentation allows for measurement of of around 750 planetary nebulae or PNe (as in the galaxy CenA, Woodley, Chakrabarty, under preparation). Such high a sample size is however more of an exception than the rule―in fact, in the real application discussed below, the number of measurements of globular clusters (or GCs) available is only 29. In addition, the measurements of are typically highly noisy, the data would typically sample the sub-space very sparsely and the data sets are typically one-time measurements. The proposed method will have to take this on board and incorporate the errors in the measurement of. Given such data, we aim to learn the gravitational mass density of all matter― dark as well as self-luminous―at any location in the galaxy.

In the Bayesian framework, we are essentially attempting to compute the posterior of the unknown gravitational mass density function, given data. Since gravitational mass density is non-negative,. That we model the mass density to depend only on location is a model assumption¹.

From Bayes rule, the posterior probability density of given data is given as proportional to the product of the prior and the likelihood function, i.e. the probability density of given the model for the unknown mass density. Now, the probability density of the data vector given the model parameters is given by the probability density function of the observable, so that, assuming the data vectors to be conditionally independent, the likelihood function is the product of the pdfs of obtained at the values of:

This is Equation (7) written in the context of this application. Given that, the pdf of is related to the pdf of the vector-valued variable as

However, this formulation still does not include the gravitational mass density function in the definition of, we explore the Physics of the situation to find how to embed into the definition of the pdf of the state space variable, and thereby into the likelihood. This is achieved by examining the time evolution of this pdf of the state space variable; we discuss this next.

We intend to make inference on each component of the vector and the matrix, along with. We do this under the constraints of a gravitational mass density function that is non-increasing with and a pdf of the state space variable that is non-increasing with. Motivation for these constraints is presented in Section 5.8. In other words, and for and. Also, here and.

First we discuss performing inference on using adaptive Metropolis-Hastings [46], while maintaining this constraint of monotonicity. We define

It is on the parameters that we make inference. Let within our inference scheme, at the n-th iteration, the current value of be. Let in this iteration, a candidate value of be proposed from the folded normal density, i.e.

where the choice of a folded normal [47] or truncated normal proposal density is preferred over a density that achieves zero probability mass at the variable value of 0. This is because there is a non-zero probability for the gravitational mass density to be zero in a given radial bin. Here and are the mean and variance of the proposal density that is proposed from. We choose the current value of as and in this adaptive inference scheme, the variance is given by the empirical variance of the chain since the -th iteration, i.e.

We choose the folded normal proposal density given its ease of computation:

It is evident that this is a symmetric proposal density. We discuss the acceptance criterion in this standard Metropolis-Hastings scheme, after discussing the proposal density of the components of the matrix and the parameter.

If is accepted, then the updated b-th component of in the n-th iteration is. If the proposed candidate is rejected then resorts back to.

Along similar lines, we make inference directly on

Let in the n-th iteration, the current value of be and the proposed value be where the

proposed candidate is sampled from the folded normal density where the variance

is again the empirical variance of the chain between the -th and the -th iteration. Then the updated general element of the state space pdf matrix in this iteration is, if the proposed value as accepted, otherwise,. Thus, the proposal density that a component of the matrix is proposed from is also symmetric.

We propose from the discrete uniform distribution, i.e. the proposed value of in the n-th iteration is

where the bounds of the interval are found experimentally given the data at hand.

Given that we are making inference on the and, we rephrase the posterior probability of the unknowns as. This posterior density is proportional to the

RHS of Equation (59).

Then given that the proposal densities that components of and of are sampled from and that the proposal density for is uniform. The Metropolis-Hastings acceptance ratio is reduced to the ratio of the posterior of the proposed state space vector value to that of the current state space vector, i.e. the proposed state space vector is accepted if

where the uniform random variable.

In this section we illustrate the methodology on synthetic data set simulated from a chosen models for the pdf of.. The chosen models for this pdf are or and. These are given by:

where with chosen in both models for the state space pdf to be. Here the model parameters and are assigned realistic numerical

values. From these 2 chosen pdfs, values of were sampled; these 2 samples constituted the 2 synthetic data sets and. The learnt gravitational mass density parameters and discretised version of the state space pdf are displayed in Figure 2. Some of the convergence characteristics of the chains are explored in Figure 3. The trace of the joint posterior probability of the unknown parameters given the data is shown along with histograms of learnt from 3 distinct parts of the chain that is run using data.

In this section we present the gravitational mass density parameters and the state space pdf parameters learnt for the real galaxy NGC3379 using 2 data sets and which respectively have sample size 164 [48] and 29 [49]. An independent test of hypothesis exercise shows that there is relatively higher support in for an isotropic pdf of the state space variable than in. Given this, some runs were performed using an isotropic model of the state space pdf; this was achieved by fixing the number of L-bins to 1. Then identically takes the value and is rendered a constant. This effectively implies that the domain of is rendered uni-dimensional, i.e. the state space pdf is then rendered. Recalling the definition of an isotropic function from Remark 5.2, we realise that the modelled state space pdf is then an isotropic function of and. Results from chains run with such an isotropic state space pdf were overplotted on results from chains run with the more relaxed version of the pdf that allows for incorporation of anisotropy; in such chain, is in fact learnt from the data.

In this work we focused on an inverse problem in which noisy and partially missing data on the measurable is used to make inference on the model parameter vector which is the discretisation of the unknown model function, where is an orthogonal transformation of and. The measurable is an unknown function of the sought function. Given that the very Physics that connects to is unknown―where―we cannot construct training data, i.e. data comprising a set of computed for a known. In the absence of training data, we are unable to learn the unknown functional relationship between data and model function, either using splines/wavelets or by modelling this unknown function with a Gaussian Process. We then perform the estimation of at chosen values of, i.e. discretise the range of values of and estimate the vector instead, where is the value of for in the -th -bin. We aim to write the posterior of given the data. The likelihood could be written as the product of the values of the of the state space vector achieved at each data point, but the data being missing, the is projected onto the space of and the likelihood is written in terms of these projections of the. is embedded within the definition of the domain of the of. The projection calls for identification of the mapping between this domain and the unobserved variables; this is an application specific task. The likelihood is convolved with the error distribution and vague but proper priors are invoked, leading to the posterior probability of the unknowns given the data. Inference is performed using adaptive MCMC. The method is used to learn the gravitational mass density of a simulated galaxy using synthetic data, as well as that in the real galaxy NGC3379, using data of 2 different kinds of galactic particles. The gravitational mass density vector estimated from the 2 independent data sets are found to be distinct.

The distribution of the gravitational mass in the system is indicated by the function.

the discretised form of this function defines the parameters,. These are computed using the learnt value of the parameters and plotted in Figure 4. We notice that the estimate of can depend on the model chosen for the state space pdf; thus, the same galaxy can be inferred to be characterised by a higher gravitational mass distribution depending on whether an isotropic state space is invoked or not. Turning this result around, one can argue that in absence of priors on how isotropic the state space of a galaxy really is, the

learnt gravitational mass density function might give an erroneous indication of how much gravitational mass there is in this galaxy and of corse how that mass is distributed. It may be remarked that in lieu of such prior knowledge about the topology of the system state space, it is best to consider the least constrained of models for the state space pdf, i.e. to consider this pdf to be dependent on both and.

It is also to be noted that the estimate for the gravitational mass density in the real galaxy NGC3379 appears to depend crucially on which data set is being implemented in the estimation exercise. It is possible that the underlying pdf of the variable is different for the sub-volume of state space that one set of data vectors are sampled from, compared to another. As these data vectors are components of and of different kinds of galactic particles, this implies that the state space pdf that the different kinds of galactic particles relax into, are different.