A dynamic unsupervised laterally connected neural network architecture for integrative pattern discovery

Abstract

We describe an unsupervised neural network approach to build associations between neurons within cortical maps. These associations are then used to capture patterns in the input data. The cortical maps are modeled using growing self-organization maps to capture the input stimuli distribution within a two dimensional neuronal map. The associations are modeled using passive lateral connections using recognition frequency of input stimuli by a neuron. The proposed approach introduces a novel way of learning by adapting neighborhood learning rules and proximity measures according to the input stimuli structure. © 2011 Springer-Verlag.