Theoretical Analysis on Classification Capability of Classifier Containing Quadratic Sigmoid Neurons
一类二次Sigmoid神经元构建的分类器分类能力的理论分析

This paper addresses the classification capability of network containing quadratic sigmoidal neurons. By analysing the size of the network required for solving Dichotomy problem, we can figure out the degree of the improvement on the classification capability of net- work by using quadratic sigmoidal neurons.