A new learning strategy of general BAMs

A new learning strategy of general BAMs

Hoa N.T., Duy B.T.

Human Machine Interaction Laboratory, Vietnam National University,

Hanoi, Viet Nam

Abstract: Bi-directional Associative Memory (BAM) is an artificial neural

network that consists of two Hopfield networks. The most important advantage

of BAM is the ability to recall a stored pattern from a noisy input, which depends

on learning process. Between two learning types of iterative learning and non￾iterative learning, the former allows better noise tolerance than the latter.

However, interactive learning BAMs take longer to learn. In this paper, we

propose a new learning strategy that assures our BAM converges in all states,

which means that our BAM recalls perfectly all learning pairs. Moreover, our

BAM learns faster, more flexibility and tolerates noise better. In order to prove

the effectiveness of the model, we have compared our model to existing ones by

theory and by experiments. © 2012 Springer-Verlag.

Author Keywords: Bi-directional Associative Memory; Hopfield neural network;

Multiple Training Strategy

Year: 2012

Source tilte: Lecture Notes in Computer Science (including subseries Lecture

Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Volume: 7376 LNAI

Page: 213 - 221

Link: http://www.scopus.com/inward/record.url?eid=2-s2.0-

84864952199&partnerID=40&md5=7e2cbe792ff2f087536af4a99690c59f

Correspondence Address: Hoa, N.T.; Human Machine Interaction Laboratory,

Vietnam National University, Hanoi, Viet Nam; email: [email protected]

Editors:

ISSN: 3029743