Topological preserving network by the existence of lateral feedback

The self-organising map can preserve the topological order of its input data. It is just a kind of competitive learning system where the neighbors of the winning neuron also participate in the learning. The structure is surprisingly simple. However, it does not consider the observation, from anatomical and physiological evidence, of the lateral feedback of neurons in the nervous system. It is found that the lateral feedback leads to the result of neighborhood learning in self-organising maps. This paper explains why the topological order of the input patterns can be captured with the existence of lateral feedback. A new activity function for each neuron during learning is proposed in order to cater for the presence of lateral feedback