Architecture Selection of Multilayered Neural Networks in Data Analysis

N. Watanabe and T. Kikuchi (Japan)


Multilayered Neural Network, Information Criterion, AIC,Cross Validation


Multilayered neural networks are used as statistical models in data analysis, because of their ability in approximation of nonlinear systems. It is an important problem to select appropriate numbers of neurons in each layer for captur ing features of data. Though this is a popular problem for statistical models, it is not easy to decide the optimal size of the neural network because of its strong nonlinear ity. We discuss problems on well-used information crite ria and propose a model selection method based on cross validation.

Important Links:

Go Back