Reza Shadmehr (John Hopkins University, School of Medicine)

The cerebellum as a neural learning machine

Begin: Wed, Jul 11, 2018 16:00

BCCN Berlin
Lecture Hall
Philippstr. 13, Haus 6
10115 Berlin

iCalendar

Abstract

The cerebellum is responsible for making accurate predictions about sensory consequences of actions. However, it has been difficult to understand how its principal cells, Purkinje cells, represent this prediction, and how that prediction is then changed following experience of error. Here, I will summarize data from recent experiments that have suggested a solution to this problem. It appears that the key is the special anatomical organization of the Purkinje cells, which appear to form small cluster, wherein the membership in the cluster is specified by the cell's preference for error. The result is an interesting internal organization of a neural network that not only learns from its prediction errors, but is empowered to reduce those errors through downstream projections onto effectors.

Organized by

Martin Rolfs/Margret Franke

← all talks