A parallel algorithm for gradient traini
β
ZdenΔk HanzΓ‘lek
π
Article
π
1998
π
Elsevier Science
π
English
β 742 KB
This paper presents a message-passing architecture simulating multilayer neural networks, adjusting its weights for each pair, consisting of an input vector and a desired output vector. First, the multilayer neural network is defined, and the difficulties arising from parallel implementation are cla