Effectiveness of Small World Network to the Performance Improvement of a Morphological Associative Memory without a Kernel Image

  • SAEKI Takashi
    Graduate School of Life Science and Systems Engineering, Kyusyu Institute of Technology
  • MIKI Tsutomu
    Graduate School of Life Science and Systems Engineering, Kyusyu Institute of Technology

Bibliographic Information

Other Title
  • 核を用いない形態学的連想記憶におけるSmall world networkの有効性
  • カク オ モチイナイ ケイタイガクテキ レンソウ キオク ニ オケル Small world network ノ ユウコウセイ

Search this article

Abstract

This article presents a new approach of the morphological associative memory (MAM) without a kernel image to improve the perfect recall rate by introducing the small world network. The MAM is one of the powerful associative memories compared to ordinary associative memories in terms of calculation amount, memory capacity and noise tolerance. However, the MAM needs the kernel image which is susceptibility to noise and difficult to design. Although, as a practical model, the MAM without a kernel image has been proposed, the model has a problem that the perfect recall rate is degraded. On the other hand, it has been reported that an introduction of the small world network to associative memories is effective in the recall rate improvement and the network size reduction. The small world network is easy to handle because it can design by β-Graph which is controlled by only one parameter. We try to improve of the perfect recall rate and reduce the network size by using the small world network. The effectiveness of the proposed approach is confirmed by autoassociation experiments.

Journal

References(22)*help

See more

Details 詳細情報について

Report a problem

Back to top