Contextual modeling using context-dependent feedforward neural nets

Abstract

The paper addresses the problem of using contextual information by neural nets solving problems of contextual nature. The models of a context-dependent neuron and a multi-layer net are recalled and supplemented by the analysis of context-dependent and hybrid nets' architecture. The context-dependent nets' properties are discussed and compared with the properties of traditional nets considering the Vapnik-Chervonenkis dimension, contextual classification and solving tasks of contextual nature. The possibilities of applications to classification and control problems are also outlined.