Realistic Modeling of Large-Scale Networks: Spatio-temporal Dynamics and Long-Term Synaptic Plasticity in the Cerebellum

A large-scale computational model of the cerebellum granular layer has been adapted to generate long-term synaptic plasticity in response to afferent mossy fiber bursts. A simple learning rule was elaborated in order to link the average granule cell depolarization to LTP and LTD. Briefly, LTP was generated for membrane potentials >-40 mV and LTD for membrane potentials <-40 mV. The result was to generate LTP and stronger excitation in the core of active clusters, which were surrounded by LTD. These changes were accompanied by a faster and stronger spike generation compared to the surround. These results reproduce the experimental observations and provide a valuable and efficient tool for implementing autonomous learning algorithms in the cerebellar neuronal network.