Ambiguously fluctuating associative memory model with hysteresis dependency

This paper proposes a new associative memory model based on an oscillatory neural network for a dynamic recognition process. A state space is partitioned by various attractor basins of stable points, limit cycles and chaos, etc., according to the initial state and property of hysteresis. The ambiguously fluctuating output is affected not only by the initial input state, but also by the prediction from the context thorough the prior input sequences.