Improved one-shot learning for feedforward associative memories with application to composite pattern association

Citation
Yq. Wu et Sn. Batalama, Improved one-shot learning for feedforward associative memories with application to composite pattern association, IEEE SYST B, 31(1), 2001, pp. 119-125
Citations number
23
Categorie Soggetti
AI Robotics and Automatic Control
Journal title
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS
ISSN journal
10834419 → ACNP
Volume
31
Issue
1
Year of publication
2001
Pages
119 - 125
Database
ISI
SICI code
1083-4419(200102)31:1<119:IOLFFA>2.0.ZU;2-5
Abstract
The local identical index (LII) associative memory (ARI) proposed by the au thors ina previous paper is a one-shot feedforward structure designed to ex hibit no spurious attractors. In this paper we relax the latter design cons traint in exchange for enlarged basins of attraction and we develop a famil y of modified LII AM networks that exhibit improved performance, particular ly in memorizing highly correlated patterns. The new algorithm meets the re quirement of no spurious attractors only in a local sense. Finally, we show that the modified LII family of networks can accommodate composite pattern s of any size by storing (memorizing) only the basic (prime) prototype patt erns, The latter property translates to low learning complexity and a simpl e network structure with significant memory savings. Simulation studies and comparisons illustrate and support the theoretical developments.