Pattern dependency of one-to-many association ability in Chaotic Complex-valued Multidirectional Associative Memory with variable scaling factor

Abstract

In this paper, we examine the pattern dependency of the one-to-many association ability in the Chaotic Complex-valued Multidirectional Associative Memory (CCMAM) with variable scaling factor. The CCMAM is composed of complex-valued neurons and chaotic complex-valued neurons, and can realize one-to-many associations of M-tuple multi-valued patterns. The one-to-many association ability of the CCMAM depends the parameters of chaotic complex-valued neurons, and we have already examined the parameter dependency of the one-to-many association ability of this model, and confirmed that the scaling factor of the refractoriness in chaotic complex-valued neurons has a significant influence on the association ability. Although it is thought that stored patterns also have influence on the association ability, the pattern dependency of the one-to-many association ability in the CCMAM has not been examined. In this research, we examined the pattern dependency of the one-to-many association ability and confirmed that the one-to-many association ability decreases when the number of stored patterns or the number of stored common patterns are large.