An Auto-Correlation Associative Memory which Has an Energy Function of Higher Order

Sadayuki MURASHIMA  Takayasu FUCHIDA  Toshihiro IDA  Takayuki TOYOHIRA  Hiromi MIYAJIMA  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E78-A   No.3   pp.424-430
Publication Date: 1995/03/25
Online ISSN: 
Print ISSN: 0916-8508
Type of Manuscript: PAPER
Category: Neural Networks
auto-correlation,  associative memory,  energy function,  higher order,  

Full Text: PDF(515.5KB)>>
Buy this Article

A noise tolerant auto-correlation associative memory is proposed. An associated energy function is formed by a multiplication of plural Hopfield's energy functions each of which includes single pattern as its energy minimum. An asynchronous optimizing algorithm of the whole energy function is also presented based on the binary neuron model. The advantages of this new associative memory are that the orthogonality relation among patterns does not need to be satisfied and each stored pattern has a large basin of attraction around itself. The computer simulations show a fairly good performance of associative memory for arbitrary pattern vectors which are not orthogonal to each other.