DMAN: distributed mass attraction network
Bildiriler > Tam Metin Bildiri
Yazarlar: M.H. Erdem, Y. Ozturk
Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94), , , Uluslararası, Tam Metin Bildiri, cilt 2, ss 1130-1136
Özet
The authors previously proposed two neural associative memory architectures, namely, mass attraction network (MAN) and equivalent mass attraction network (EMAN), based on Newton's mass attraction theory. DMAN is a neural network model based on the same concepts, but unlike MAN or EMAN, which use local representations, DMAN uses distributed representations. After introducing some general issues, this paper explains DMAN concept and the network. Computer simulations revealed much better performance than the Hopfield model on the basis of the capacity of the model and error.>