Self-Organization and Entropy Decreasing in Neural Networks
Masayoshi Inoue, Masayuki Kashima · Progress of Theoretical Physics · 1994
Dynamics of self-organization of binary patterns in a Hopfield model, a Boltzmann machine and a chaos neural network are investigated with the use of an ensemble average entropy and a short time average entropy. Time dependences of these entropies are calculated by numerical simulations when these models are solving traveling salesman problems. Decreasing of the entropies are observed in consequences of the self-organization at their initial time stages.