École Nationale Supérieure de
Techniques Avancées

Abstract

In the field of neuroscience, the continuous creation of new, larger, and more complex datasets presents challenges in extracting meaningful information. This is often exacerbated by the "black box" nature of certain machine learning algorithms and the entanglement of data. This report explores into the implementation of a modified Restricted Boltzmann Machine model introduced by Fernandez-de-Cossio-Diaz and others. in their article "Disentangling Representations in Restricted Boltzmann Machines without Adversaries" [7], with neural data. The report begins with an introduction to fundamental concepts, followed by initial testing of the model in its simplest form, and the presentation of the obtained results. Subsequently, new configurations are explored to uncover potential ways of implementing this modified RBM and finally the results are compared with an other dataset.