Probability limit property for energy function to feed-forward neural networks with noise
Cong Jin · 2003
A probability limit property is proposed for the weight vectors W of feed-forward neural networks when both the input data and output data contain noise or when only the output data contains noise. By theoretical analysis of the energy function of a feed-forward neural network, the paper points out that a least square energy function isn't a good choice. The result is good enough for future research.