Output details
11 - Computer Science and Informatics
University of Surrey
Classification of Distorted Patterns by Feed-Forward Spiking Neural Networks
<24>Natural neural systems are noise-tolerant, and so must be neural learning models. This paper addresses the question of whether "meaningful" spatio-temporal spike patterns are recognised under addition or omission of spikes in a spike pattern _mapping_ task. Our approach is natural, yet novel in that existing approaches have used less-distorting spike jitter as the source of noise. A network performing a mapping task produces a particular spike train in response to its input. Hence our approach forms the basis for _robust_ models of neural processing of a-priori unrelated spatio-temporal pattern pairs, such as transforming sensoric input to particular motoric reaction.