Data fusion by joint non-negative matrix factorization for hypothesizing pseudo-chemistry using Bayesian networks

Abstract

<p>Inferring the reaction pathways underlying the processing of complex feeds, using noisy data from spectral sensors that may contain information regarding molecular mechanisms, is challenging. This is tackled by a...</p>

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks

“…The discussion of the tensor decomposition and the subsequent interpretation of the Bayesian networks constructed from the pseudo-component spectra are provided together with the results for each case so as to have an easier interpretation. The reaction pathways hypothesized from the Bayesian networks have been validated against the literature pertaining to conversion chemistry in bitumen that has been investigated using quantitative metrics reflecting composition changes of model compounds, representative of the complex reactive system. − , It must be noted that the pseudo-component signatures from the tensor decomposition do not point to a single molecular structure but a class of compounds. Suitable model compounds with structures representative of the pseudo-component spectra have been used to indicate plausible conversion pathways in line with the Bayesian networks.…”

“…In our earlier work, where joint non-negative matrix factorization had been demonstrated as a data fusion algorithm to extract pseudo-component spectra, the number of components had been empirically determined using the notion of chemical rank . The initially obtained reaction hypotheses were seen to indicate the cracking and hydrogen transfer in substituted aromatics as well as esterification to produce condensed aromatics and esters.…”

“…The latent factor projections in the feature space of spectral channels of the sensors gain interpretability as the pseudo-component spectra from which the dominant compound classes underlying the complex feed are deduced using expert knowledge. These compound signatures are then represented as nodes among which the causal paths are learned using Bayesian structure learning as a way of generating reaction hypotheses . The hypotheses are then validated using domain knowledge to infer the chemical interactions among the identified compound classes.…”

“…The multi-modal structure of the tensor data from both sensors is exploited by their low rank representation to yield latent factors across the unshared mode of spectral channels, physically interpreted as pseudo-component spectra for each of the sensors in accordance with Beer’s law . The interactions among the pseudo-components are proposed to be encoded by a causal framework in an attempt to build an inferential model to hypothesize the underlying pseudo-reaction chemistry during the upgrading process. , …”

Estimating reaction parameters in mechanism‐enabled population balance models of nanoparticle size distributions: A Bayesian inverse problem approach

“…https://orcid.org/0000-0003-2311-9402 ENDNOTES * In our literature search, we found a number of uses of Bayesian inversion in chemistry. 18,[43][44][45][46][47][48][49][52][53][54][55][56][57][58][59][60][61][62][63][64] Armstrong and Hibbert 50,51 also provide a comprehensive, albeit now decade-old survey of the uses of Bayesian methods in chemistry. However, our attempt at a comprehensive literature search revealed no uses of Bayesian methods for nanoparticle mechanistic chemistry nor evidence for its deserved, more extensive use in mechanistic chemistry in general.…”