Title :
Parallel neurocomputer realisation of the backpropagation algorithm
Author_Institution :
Res. Inst. for Multiprocessor Comput. Syst., Taganrog, Russia
Abstract :
The authors discusses the possibilities of the parallel digital neurocomputer realization of the backpropagation neuronlike networks i.e., their PDNPA (parallel digital neurocomputer with programmable architecture)-based realization based on the neurocomputing performed by the DNP. The results obtained make it possible to construct parallel digital neurocomputers taught by the backpropagation method, which can provide an effective solution for a wide class of pattern recognition problems in real time
Keywords :
backpropagation; neural nets; parallel architectures; backpropagation neuronlike networks; parallel digital neurocomputer realization; programmable architecture; Approximation algorithms; Backpropagation algorithms; Multi-layer neural network; Neural networks;
Conference_Titel :
Neuroinformatics and Neurocomputers, 1992., RNNS/IEEE Symposium on
Conference_Location :
Rostov-on-Don
Print_ISBN :
0-7803-0809-3
DOI :
10.1109/RNNS.1992.268578
