论文信息 - Special Statistical Properties of Neural Network Learning

Special Statistical Properties of Neural Network Learning

| We elucidate essential di erences between feed-forward neural network models and conventional linear statistical models. When the target is overrealizable, the MLE of the former shows worse generalization, while experimental results reveals that iterative learning of a neural network shows eminent overtraining and better generalization in the middle.

K. Fukumizu

[1] M. Opper. Learning in Neural Networks: Solvable Dynamics , 1989 .

[2] Kurt Hornik,et al. Learning in linear neural networks: a survey , 1995, IEEE Trans. Neural Networks.

[3] 福水健次. A regularity condition of the information matrix of a multilayer perceptron network , 1996 .

[4] Kenji Fukumizu,et al. A Regularity Condition of the Information Matrix of a Multilayer Perceptron Network , 1996, Neural Networks.