A Modified Bidirectional Decoding Strategy Based on the BAM Structure

Wen June Wang, Donq Liang Le

Research output: Contribution to journalArticlepeer-review

9 Scopus citations

Abstract

Based on associative memories (BAM) strategy, a modified bidirectional decoding strategy (MBDS) is introduced. With the aid of the augmented correlation matrix M = [M\My]( or M = [MT| Mx]), the long term memory (LTM) traces Ty (or Tx), and the activating function Gy(·) (or G*(·)), we construct the MBDS such that all given training pairs are guaranteed to be recalled successfully. In this paper, we also illustrate some examples for the comparison between the dummy augmentation method of Wang et at., and our MBDS. It is obvious that the MBDS save much more extra connecting weights than the dummy augmentation method.

Original languageEnglish
Pages (from-to)710-717
Number of pages8
JournalIEEE Transactions on Neural Networks
Volume4
Issue number4
DOIs
StatePublished - Jul 1993

Fingerprint

Dive into the research topics of 'A Modified Bidirectional Decoding Strategy Based on the BAM Structure'. Together they form a unique fingerprint.

Cite this