A learning scheme for bipartite recurrent networks and its performance

Author

Kumazawa, Itsuo ; Fukuda, Mitsuyoshi

Author_Institution

Dept. of Comput. Sci., Tokyo Inst. of Technol., Japan

fYear

1993

fDate

24-26 Nov 1993

Firstpage

34

Lastpage

37

Abstract

A new learning scheme specialized to recurrent neural networks with the bipartite topology is proposed. The scheme is expected to have better convergence than the general Boltzmann machine learning. This improvement results from the restricted form of the network topology and an energy form devised to have a dominant global minimum. Compared to the recurrent backpropagation algorithm, the scheme is simple, more suitable for hardware realization and its probabilistic nature reduces the effect of spurious local minima. The performance of the scheme is partly demonstrated by simulations of associative memory and compared with the general Boltzmann machine learning

Keywords

Boltzmann machines; backpropagation; content-addressable storage; learning (artificial intelligence); recurrent neural nets; uncertainty handling; Boltzmann machine learning; associative memory; bipartite recurrent networks; bipartite topology; convergence; dominant global minimum; energy form; hardware realization; learning scheme; network topology; probabilistic nature; recurrent backpropagation algorithm; recurrent neural networks; simulations; Associative memory; Backpropagation; Computer science; Machine learning; Network topology; Neural networks; Optical computing; Optical fiber networks; Optical interconnections; Optical network units;

fLanguage

English

Publisher

ieee

Conference_Titel

Artificial Neural Networks and Expert Systems, 1993. Proceedings., First New Zealand International Two-Stream Conference on

Conference_Location

Dunedin

Print_ISBN

0-8186-4260-2

Type

conf

DOI

10.1109/ANNES.1993.323088

Filename

323088