MEMORY ENCODING BY OSCILLATOR DEATH

We study associative-memory properties of networks of neural oscillators which are described by the phase degrees of freedom (phase-rotators). It is shown that a certain network of phase rotators exhibits an associative-memory function created by the cessation of oscillations rather than the synchronization. The phenomenon corresponds to the memory retrieval by oscillator death. The equilibrium properties of the model network are both analytically and numerically investigated using the self-consistent signal-to-noise analysis which was proposed for the analogue-neuron networks with a wide class of input-output functions of a neuron. Besides the standard type of storage capacity which ensures the presence of retrieval solutions, there exists another critical storage level below which the network ceases to have perfect cessation of oscillations in retrieval. A similar critical storage level is obtained for spin glass solutions.