The rockefeller university

Event Details

Type

Special Seminar Series

Speaker(s)

Srdjan Ostojic, Ph.D., director of research, French National Center for Scientific Research (CNRS), École Normale Supérieure Paris

Speaker bio(s)

Unraveling the relation between the structure of networks of neurons and the computations they perform is one of the central goals of neuroscience. Recurrent neural networks trained to perform cognitive tasks have become an influential class of models to address this question. Such models provide fully observable and functional systems, yet using them to identify general principles that relate structure and function has so far been challenging. In this seminar, Dr. Ostojic will review novel approaches developed for extracting computational mechanisms from trained recurrent networks and show how they reveal complementary functional roles for two types of connectivity structure: feedback loops and cell classes.

The overarching aim of Dr. Srdjan Ostojic’s research is to develop a mechanistic understanding of how thousands of neurons in the brain work together to implement computations that underlie behavior. To this end, he studies models of networks of neurons by combining methods from machine learning with mathematical tools inspired by statistical physics. After earning an M.S. in Physics at Ecole Polytechnique Federale Lausanne in 2002, Dr. Ostojic obtained a Ph.D. in Physics at the Universiteit van Amsterdam in 2006. He progressed through roles as a Postdoctoral Fellow in Neuroscience at Ecole Normale Supérieure Paris (2009) and Columbia University (2011), a visiting professor at both MIT and Princeton, and is currently a Research Director (DR2 CNRS) at The Laboratoire de Neurosciences Cognitives at Ecole Normale Supérieure Paris.

This Special Seminar will take place in Carson Family Auditorium and virtually via Zoom. We recommend virtual participants log out of VPN prior to logging in to Zoom. Please do not share the link or post on social media. 

Open to

Campus Only