Improved Pseudo-Relaxation Learning Algorithm for Robust Bidirectional Associative Memory
In this paper, we propose Improved Pseudo-Relaxation Learning Algorithm for Bidirectional Associative Memory (IPRLAB). Since the proposed IPRLAB is based on the conventional PRLAB, it can guarantee the recall of all training pairs and has high storage capacity. Furthermore, the proposed IPRLAB can much improve the noise reduction effect of the BAM and contribute to construct a robust memory. A number of computer simulation results show the effectiveness of the proposed learning algorithm.
KeywordsTraining Data Solution Space Linear Inequality Recall Rate Direction Cosine
Unable to display preview. Download preview PDF.
- Hagiwara, M.: Multidirectional Associative Memory. Proc. of IJCNN′90, Washington, D.C., 1, pp.3–6 (1990).Google Scholar
- Morita, M., Yoshizawa, S. and Nakano, K.: Analysis and Improvement of the Dynamics of Autocorrelation Associative Memory. IEICE Japan, J73-D-II, no.2, pp.232–242 (1990).Google Scholar
- Hattori, M., Hagiwara, M. and Nakagawa, M.: Quick Learning for Bidirectional Associative Memory. IEICE Japan, E77-D, no.4, pp.385–392 (1994).Google Scholar
- Hattori, M. and Hagiwara, M.: Intersection Learning for Bidirectional Associative Memory. Proc. of ICNN′96, 1, pp.555–560 (1996).Google Scholar
- Hattori, M. and Hagiwara, M.: Neural Associative Memory for Intelligent Information Processing. Proc. of KES′98, Adelaide, 2, pp.377–386 (1998).Google Scholar