SIMULATION OF LEARNING-PROCESSES IN NEURONAL NETWORKS OF THE CEREBELLUM

A dynamic model of learning based on the specific neuronal system of the cerebellum, including some of its structural-functional peculiarities, is proposed. It allows simulation modification processes of the parallel fiber synapse that influences the Purkinje cell. Regularities of synaptic modifications are obtained by extrapolating well-known experimental data about changes of synaptic efficiency as resulting from release, refilling and mobilization of the mediator. A mathematical description of synaptic processes corresponds to experimental data on the changes of synaptic efficiency under rhythmical stimulation and gives objective quantitative estimates for long-term (refilling of the mediator) and short-term (mobilization of the mediator) effects which are caused by presynaptic stimulation. Computer simulations were conducted to investigate the characteristics of learning for different values of the following parameters: intensity of unconditioned stimulus (US, activity of the climbing fiber), intensity of conditioned stimulus (CS, activity of the parallel fiber), temporal shift between US and CS, temporal interval between reinforcements. The temporal shift between CS and US is one of the major factors that influence the learning procedures. Analysis of the data obtained shows that the model enables simulation of the main regularities of establishment and extinction of conditioned reflexes.