A modified asynchronous chaotic neural network model for VLSI implementation

An asynchronous chaotic neural network model uses spatio-temporal patterns of neural activities for computation. The output of the neuron in the model is an impulse with an analog amplitude. The neuron utilizes the time relations among incoming pulses for information processing. In this paper, the model is modified by using a pulse with finite width instead of the impulse with infinitesimal width in order to implement the model as an integrated circuit form. It is shown that the amplitude of the internal state of the modified model doesn't show a chaotic behavior unlike the original model. However, the intervals of the output pulses express a complex behavior. Furthermore, it is confirmed that the information processing ability of the modified model is equivalent to that of the original model.