Adapting to changes in the environment is a fundamental aspect of learning. The cerebellum is widely known for its role in motor adaptation, and its highly regular circuitry has long motivated theories of pattern separation and high-dimensional representations. However, it has remained challenging to relate these classic ideas to how the cerebellum coordinates with other brain regions during complex behaviors. I will present our work testing and extending these theories. First, using large-scale recordings from hundreds of granule cells in spontaneously behaving animals, we show that population activity is high-dimensional, consistent with the idea that the cerebellum provides a rich basis for sensorimotor learning. Next, I will present theoretical work extending these principles to distributed learning across brain regions. We propose a model in which adaptation emerges from interactions between two systems: the cerebellum, acting as a fast feedforward “adapter” that responds to perturbations, and the motor cortex, acting as a slow recurrent “controller” that stores stable memories. In this framework, supervised learning in the cerebellum generates predictive error signals that guide consolidation in the motor cortex through a local plasticity rule. Together, this work helps to link high-dimensional cerebellar representations to coordinated learning across the brain during complex motor behaviors.
Organizer
Ellen Boven
e.boven@erasmusmc.nl