Classification properties and classification mechanisms of feedforward neural network classifiers

This paper studies the classification properties and classification mechanisms of outer-supervised feedforward neural network classifiers (FNNC). It is shown that nonlinear FNNCs can break through the "bottleneck" behaviours for linear FNNCs. Assume that the involved FNNCs are classifiers that associate only one output node with each class , after the global minimum solutions with null costs based on batch-style learning are obtained , it is shown that in the case of the linear outputed network classifiers , the class weight vectors corresponding to different output nodes are orthogonal, and in the case of sigmoid output acviation functions, the jth class weight vector must be situated in the negative direction of the i(i not equal j) th class weight vector.