Fuzzy-like Decision Making in a Neural Network with Chaotic Neurons and Dynamic Synapses

• 作 者: ZhijieWang;HongFan;
• 机 构: College of Information Science and Technology, Donghua University, 200051, Shanghai, China
• 出 处: the 2nd Shanghai International Symposium on NOnlinear Scienc
• 发表日期: 2005年期

　　It is well known that an autoassociative neural network can store memories through its synaptic connections [3][4].Unlike a conventional autoassociative neural network which evolves to settle at a stable steady state,a chaotic associative neural network generates peculiar behavior that the network retrieves some stored patterns and escapes from them in the transient phase [1][2][5].Such a chaotic associative memory is constructed with chaotic neuron models interconnected through a conventional auto-associative matrix of synaptic weights.On the other hand,once the weights of the synapses in a conventional autoassociative neural network are learned by learning rules,they are fixed during the memory retrieval process.This implies that the synapses of the neurons are assumed to be ‘static’,i.e.,that they change their weights only on the slow time scale.However,it has been discovered that synaptic plasticity occurs across many time scales-from the order of days to the order of milliseconds [8].The dynamic association in a neural network when a synaptic depression is included in the network is investigated [7].Similar to a chaotic associative memory,the output of such a network transits among the stored patterns during the retrieval process.