SELF-LEARNING OF PROBABILITY-DISTRIBUTION FUNCTION BY MULTILAYERED PERCEPTRONS

Overlearning in the neural network can cause a failure in the approxim ation of a function because of the limited generalization power of the neural network. To solve this problem, it is considered effective tha t the neural network learns the probability density function of the le arning samples, and the result ''answer is impossible'' is obtained wh en an input is given in the region containing no learning sample. This paper proposes a method in which the probability density of the learn ing sample is reflected on the output for the function by alternately inputting the learning sample and the random value. It is reported in this paper through a simulation that the proposed method works correct ly in the multilayered perceptron (MLP). When this method is applied, there is a close relation among the SN ratio, the output norm and the probability density of the learning samples.