Mitsuo KAWATO and Hiroaki GOMI
A Computational Model of Four Regions
of the Cerebellum
Based on Feedback-Error-Learning
Abstract:We propose computationally coherent models of the four regions of the cerebellum
based on the feedback-error-learning scheme. We assume that climbing fiber
responses represent the efference copy of motor commands generated by premotor
networks such as feedback controllers at the spinal-, brain stem- and cerebral levels.
Based on the long-term depression in Purkinje cells, each corticonuclear microcomplex
in different regions of the cerebellum learns to execute predictive and coordinative
control of different types of movements. Ultimately, it acquires an inverse
model of a specific controlled object and complements crude control by the premotor
networks. Thus, premotor network activity decreases as the learning proceeds.
As a representative example, computer simulation of simultaneous adaptation of the
vestibulo-ocular reflex and the optokinetic eye movement response was successfully
performed while the Purkinje cells receive eye-velocity signal by recurrent neural
connections as well as the vestibular input and the retinal slip as parallel fiber inputs.