Séminaire ayant lieu dans le cadre du doctorat en informatique cognitive, en collaboration avec le centre de recherche CRIA  

TITRE :  Cortical feedback mechanisms in visual reasoning: From perceptual grouping to abstract compositional reasoning

Thomas SERRE

Jeudi le 12 février 2026 à 10h30

Local PK-5115 (Il est possible d'y assister en virtuel en vous inscrivant ici)              

RÉSUMÉ

This talk examines how cortical feedback facilitates compositional visual reasoning, distinguishing biological from artificial vision. Same-different judgments—fundamental symbolic operations that newborn ducklings master from single examples yet challenge state-of-the-art feedforward networks—illustrate this gap. Our computational studies suggest that structured, object-centered representations enable efficient learning of such abstract relations. These tasks require attention and working memory—inherently recurrent processes—as confirmed through human neurophysiological recordings. How are such structured representations built? Compositional understanding—from feature binding to relational and abstract reasoning—depends on cortical feedback. Brain-inspired recurrent models tuse feedback for iterative refinement of compositional representations. These solve challenging visual reasoning tasks—from curve tracing to object tracking—that remain difficult for feedforward architectures, including transformers. Recent evidence that primates use visual mental simulation when solving complex problems further supports the computational role of feedback in maintaining internal representations during reasoning. Together, these findings provide computational evidence that cortical feedback contributes essential mechanisms for the compositional reasoning capabilities that connect perception and abstract thought.

BIOGRAPHIE

Thomas SERRE is the Thomas J. Watson, Sr. Professor of Science and Professor of Cognitive & Psychological Sciences and Computer Science at Brown University. He is the Faculty Director of the Center for Computation and Visualization and Associate Director of the Center for Computational Brain Science. He also holds an International Chair in AI in the Artificial and Natural Intelligence Toulouse Institute (France). His research focuses on understanding neural computations supporting visual perception, with particular emphasis on the role of recurrent and feedback processes in visual reasoning.

RÉFÉRENCES

 Muzellec, D. Linsley, A.K. Ashok, E. Mingolla, G. Malik, R. VanRullen & T. Serre. Tracking objects that change in appearance with phase synchrony. International Conference on Learning Representations, 2025

A. Ahuja, N.Y. Rodriguez, A.K. Ashok, T. Serre, T. Desrochers & D. Sheinberg. Monkeys engage in visual simulation to solve complex problems. Current Biology, 2024

A. Alamia, C. Luo, M. Ricci, J. Kim, T. Serre & R. VanRullen. Differential involvement of EEG oscillatory components in sameness vs. spatial-relation visual reasoning tasks, eNeuro, 2020

M. Ricci, R. Cadene & T. Serre. Same-different conceptualization: A machine vision perspective. Current Opinion in Behavioral Sciences, 2020