Evolution of Associative Memory with Environmental Change

There have been a lot of researches which apply evolutionary techniques to layered neural networks. However, their applications to Hop eld neural networks remain few so far. We have been applying a genetic algorithm to fully connected associative memory model of Hop eld, and reported elsewhere that the network can store some number of patterns only by evolving weight matrices with the genetic algorithm. In this paper, we present that evolving both weight matrices and patterns to be stored results in the higher capacity of associative memory than in the case of xed patterns above. From an Arti cial Life perspective, the patterns to be stored are considered to be an environment.