A Chaos Association Model with a Time-Dependent Periodic Activation Function

摘要

In this paper we shall propsoe an autoassociation model with a parameter controlled periodic activation function. In practice the activation function will be controlled between a monotonous periodic function and a nonmonotonous one to acheive a high ememory capacity avoiding an unfavourable trapping at a spurious state. The presently proposed model involves a chaotic dynamics in itself and realises a chaotic wandering mode as well as an autoassociation mode. In practice, as a result of the computer simulation, one sees that the complete association can be realised up to the loading rate L/N approx 0.5 even for the searching mode.