Title: Structured Recurrent Networks: Two Lessons from Random Matrix Theory
Speaker: Yonatan Aljadeff (Postdoctoral Scholar, University of Chicago)
Abstract: Recently we developed a dynamic mean field approach for partially structured and partially random neural networks. These networks undergo a phase transition from a single fixed point to chaos at a critical point related to a new result in mathematics – the spectral radius of the connectivity matrix. I will present these advances and discuss the insight they give into the questions: what could cell-types (~modularity) be good for? and should we invest in acquisition of large scale connectomes?
Joint work with D. Renfrew, T. Sharpee, M. Stern, M. Vegué
This seminar is co-sponsored by Case Western Reserve University’s Institute for the Science of Origins (ISO).