Title :
A virtual network for reorganizable neural architectures
Author :
Wong, C.Y. ; Lursinsap, C.
Author_Institution :
Center for Adv. Comput. Studies, Univ. of Southwestern Louisiana, Lafayette, LA, USA
Abstract :
Summary form only given. The authors propose a hardware-implementable, virtually connected, reconfigurable network architecture that has a high degree of fault tolerance. This architecture provides the neurons the ability of self-organization, and a neuron can establish a new synaptic connection that can be either excitatory or inhibitory, replacing a weakened old synapse. Cells are grouped into small partitions called cell cluster, and the communications among cells are limited within their influence regions. A message broadcasting algorithm which effectively limits the reach of the cells within their influence regions is described, and some illustrations of the fault tolerance characteristic of the architecture are given. By combining the cell clusters with the communication algorithm, a network with virtually no limit in size can be constructed.<>
Keywords :
fault tolerant computing; neural nets; parallel architectures; virtual machines; cell cluster; fault tolerance; message broadcasting algorithm; neural nets; reorganizable neural architectures; synaptic connection; virtual network; Computer fault tolerance; Neural networks; Parallel architectures; Virtual computers;
Conference_Titel :
Neural Networks, 1989. IJCNN., International Joint Conference on
Conference_Location :
Washington, DC, USA
DOI :
10.1109/IJCNN.1989.118402