A Study of Chaos Synergetic Neural Network(Session II : Chaos, The 1st Tohwa University International Meeting on Statistical Physics Theories, Experiments and Computer Simulations)
Masahiro Marshall Nakagawa · Institutional Repositories DataBase (IRDB) · 1996
In this work we shall put forward a novel synergetic neural network involving a chaos dynamics and investigate the dynamic properties in memory retrieval process. The present artificial neuron model is characterized by a sinusoidal activation function as well as competitive nonlinear connections between synergetic neurons. It is elucidated that the present neural network has an advantage in the dynamic memory retrievals beyond the conventional chaotic model with such a monotonous mapping as a sigmoid function. This finding is considered to result from both the nonmonotonous property of the analogue periodic mapping which may be accompanied with a chaotic behaviour of the neurons and the synergetic connections related to the competition between the overlaps. It is also shown that the present analogue neural network may be reduced to the previously proposed synergetic neural network if one assumes an identity mapping in the feedback loop instead of the periodic one.