| old_uid | 19860 |
|---|---|
| title | titre non précisé |
| start_date | 2021/12/13 |
| schedule | 11h30 |
| online | no |
| location_info | Amphithéâtre |
| details | Chaired by Henry Kennedy |
| summary | The neocortex uses slow processing elements to solve in near real-time a va- riety of computational problems that cannot yet be satisfied by computers that have very much faster hardware. Understanding the architecture and princi- ples of processing of cortical circuits would open the way to explaining thought, as well as to the development of novel computing technologies. Thirty years ago Kevan Martin, David Whitteridge and I reported that characteristic in- tracellular signals derived from different types of cortical neurons in vivo are consistent with their participation in a ’canonical’ circuit that is dominated by recurrent excitation amongst its superficial pyramidal cells (Neural Computa- tion, 1989). In this talk I will revisit those findings, and then describe a family of neuronal network models that make use of recurrence for signal enhance- ment, state-dependent processing, and constraint satisfaction. I will argue that it is the unstable dynamics inherent in positive feedback that drives cortical computation. Douglas, D.J, Martin, K.A.C., and Whitteridge, D. A Canonical Microcircuit for Neocortex (1989) Neural Computation 1, 480-488. |
| responsibles | Epinat-Duclos |