Oscillatory neural network for dynamical associative memory

摘要

A discrete time oscillatory neural network (DTONN) for the dynamical associative memory is proposed in this paper. The network reveals fruitful dynamical behaviors when applied to associative memory for hand-written character patterns. It can recall the stored non-orthogonal patterns from the network when it is activated by an initial pattern contaminated by noise. The network also demonstrates the synchronous oscillations and spatio-temporal chaos during the auto-evolutionary procedure with time.