Ensemble coding in neural network system learning by negative reinforcement

A biologically plausible model of the system with an adaptive behavior in an a priori uncertain environment and resistance to impairment has been developed. The system consists of input (sensory), learning, and output (motor control) subsystems. The first one classifies input patterns presented as n-dimensional vectors in accordance with some associative rule. The second subsystem being a neural network with mutual inhibitory connections determines adaptive responses of the system to input patterns during learning by negative reinforcement. Arranged neural ensembles coding appropriate output responses are formed during learning. The output subsystem maps a neural network state into the system behavior in the environment. The system developed has been studied by computer simulation imitating a collision avoidance by a mobile robot. After some learning period, the system “moves” along a road without collisions with obstacles. It is shown that in spite of impairment of some neural network elements the system functions reliably after relearning