Description
Date depot: 12 mars 2024
Titre: Emergence of mesoscale properties in neural networks
Directeur de thèse:
Fabrizio DE VICO FALLANI (ICM)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Sciences de l’information et sciences du vivant
Resumé: Current artificial intelligence (AI) achieve unbeatable performance in solving specific
tasks. However, they are still far from being versatile and generalizable, a basic feature of many living
organisms. Recent evidence indicates that mesoscale properties of brain networks, such as structural
modularity, are crucial for cognitive flexibility. While mesoscale properties have been proven to en-
hance task-based performance, their role in generalizable AI is still poorly understood. This project
aims to elucidate the role of mesoscale structures in generalizable AIs inspired by the development of
real brain networks. By considering synthetic models of growing networks, we first aim to characterize
the emergence of mesoscale properties from a theoretical perspective. The parameters of the model
will be then fine-tuned considering the brain wiring formation in different species. These models will
be eventually used to design a new generation of artificial neural networks and test their generalizability.
This project bridges the gap between machine learning and network science, with potential
applications in both artificial and biological intelligence.