Exploring the role of neural network architecture onto decision-making processes with reservoir computing.

novembre 2024

Directeur(s) de thèse : Nicolas Rougier and Thomas Boraud

Résumé de thèse

A striking similarity exists across diverse species in the structure of certain brain regions, strongly suggesting that they have been conserved throughout evolution. This observation raises fundamental questions about the origins of these structures and the potential for alternative ones to achieve similar cognitive functions. To address this question, this thesis explores the relationship between brain architecture and cognitive function, with a focus on decision-making processes. We use variants of a recurrent neural network model that is structurally minimal. The latter successfully solves simple decision tasks; however, when applied to more complex tasks, the model encounters performance limitations, requiring additional structural constraints to achieve better results. We subsequently employ a genetic algorithm to evolve network structure to more complex ones, leading us to discover multiple realizable solutions emerging through structural variations. Furthermore, our findings reveal that identical architectures can exhibit a range of different behaviors, leading us to investigate additional factors contributing to these behavioral differences beyond structural variations. Our analysis of the behavior of 24 monkeys living in a community reveals that social factors, such as social hierarchy, play a significant role in influencing behavior.

This thesis takes an approach that differs from traditional neuroscience methodologies. Rather than directly constructing biologically inspired architectures, the models are designed from simple to complex structures, reproducing the process of biological evolution. By leveraging the principles of multiple realizability, this approach enables the evolution of diverse structural configurations that can achieve equivalent functional outcomes.