Variable-sized KFM associative memory with refractoriness based on area representation

In this paper, we propose a variable-sized KFM associative memory with refractoriness based on area representation. In the proposed model, the connection weight fixed and semi-fixed neurons are introduced, and the pattern that has already been learned is not destroyed and a new pattern can be memorized. Moreover, when unknown patterns are given, neurons can be added in the map layer if necessary. We carried out a series of computer experiments, and confirmed that the proposed model can learn new patterns which has one-to-many relations successively, neurons can be added in the map layer if necessary, and the proposed model has robustness for noisy input and damaged neurons.