Cerebral mechanism for reward-mediated learning: A mathematical model of neuropopulational network plasticity (Q922380)

A formal mathematical description of a neurophysiological approach to learning by reward is presented. To achieve the learning, the brain has to find rewarding actions out of a variety of (motoric) trials which would bring reward (like rats pressing a pedal). The authors suppose that this function requires two types of memory: short-term memory (STM) storing generated trials and long-term memory (LTM) which records rewarding candidates from the trials. The formal learning model is based on the post-tetanic potentiation and catecholamine dependent plasticities formulated according to physiological data. The computational model of the cerebral network includes only the brain architecture (thalamocortical system) that produces necessary and sufficient functions for the learning ability. Computer simulations are analyzed and show the way how to design more feasible experiments to verify the brain mechanisms for learning.