An artificial network based on unfixed states number oscillators

The properties of neural networks based on different oscillators with an unfixed number of states and intended for optical and hybrid implementation are investigated. The pattern recognition time depends on the computation accuracy and floating threshold, and may be less than the oscillation period. A model of the network has been developed which makes it possible to investigate some network features for different parameters and functions, and to describe the oscillator interaction methods, the learning rules, the pattern correlation indicators, and the synchronization level indexes