Summary
the 2014 International Symposium on Nonlinear Theory and its Applications
2014
Session Number:C2L-C
Session:
Number:C2L-C3
Performance analysis of Bidirectional Associative Memories by using the Inverse Function Delayless model
Chunyu Bao, Takeshi Onomi, Yoshihiro Hayakawa, Shigeo Sato, Koji Nakajima,
pp.516-519
Publication Date:2014/9/14
Online ISSN:2188-5079
Summary:
There are some problems in the conventional Hopfield model such as spurious memory and the presence of a non-optimal solution due to local minimum state. The Inverse function Delayed (ID) model has been proposed by Nakajima[5] as a neuron model. The ID model can solve such problems at 100% rate by its negative resistance effect in its dynamics, which can destabilize undesirable states selectively by this effect. However, in the neural network made by ID neurons, we can not solve the problems that have big size in real-time because of the high cost of the calculation time of the ID model. In the Inverse Function Delayless model, we proved that such problem can be avoided by realizing the destabilized region due to a discrete operation. In this report, I attempt to introduce the effect of IDL to the associative memory system of hetero associative type. However, it was associated systems BAM type in order to fulfill the effects of the IDL. In the of the proposed system, we can expect the expansion of the Basin size and speed of convergence time to the memory state.