March 18, 1994
Dimitris Stassinopoulos and Per Bak
Brookhaven National Laboratory Department of Physics Upton, NY11973, USA
The Niels Bohr Institute Department of Physics Copenhagen2100, Denmark
We have constructed a simple model of the brain. The model consists of a set of randomly connected neurons. Inputs and outputs are also connected randomly to a subset of neurons. For each input there is a set of output neurons which must fire in order to achieve success. A signal giving information as to whether or not the action was successful is fed back to the brain from the environment. The connections between firing neurons are strengthened or weakened according to whether or not the action was successful. The system learns, through a self-organization process, to react intelligently to input signals, i.e., it learns to quickly select the correct output for each input. If part of the network is damaged, the system relearns the correct response after a training period. We simulated Pavlov's experiment by suddenly changing the correct response for a certain input signal. The system continues its old response for an extended period until it eventually learns the new correct response.