Neural networks and belief logic

Abstract

Many researchers have observed that neurons process information in an imprecise manner - if a logical inference emerges from neural computation, it is inexact at best. Thus, there must be a profound relationship between belief logic and neural networks. In Chen (2002), a plausible neural network model that can compute probabilistic and possibilistic logic was proposed. In this article we further extend this model to continuous variables for function and relation estimation. We discuss why and how belief logic is derived from neural computation.