E. Del Moral Hernandez (Brazil)
chaotic neural networks, bifurcating recursive processing elements, bifurcation neurons, associative memories, artificial neural networks, bifurcation and chaos, attractor networks
This paper addresses neural architectures based on chaotic nodes, for which the local dynamics is expressed by bifurcating recursions. More specifically, it discusses the effectiveness of using different strategies for the coupling between the network's nodes. This study is done considering an example application: the storage and recovery of binary strings in associative memory architectures. The choice of coupling strategy has direct impact on: (a) the behavior of the network during the search phase, i.e., when the network is looking for a (spatio-temporal) collective pattern that is compatible with one of the stored memories; (b) the level of performance of the associative architecture.
Important Links:
Go Back
