Title :
A choosing method of training sets which prevent over-learning
Author :
Yamasaki, Kazutaka ; Ogawa, Hidemitsu
Author_Institution :
Dept. of Comput. Sci., Tokyo Inst. of Technol., Japan
fDate :
27 Jun-2 Jul 1994
Abstract :
We discuss the over-learning problem for feedforward neural networks. We have previously given the necessary and sufficient conditions for causing over-learning. In this paper we discuss, based on these conditions, a method for choosing the training sets. As an example, we give a method for choosing the training sets so that the memorization learning does not cause Wiener-over-learning
Keywords :
feedforward neural nets; inverse problems; learning (artificial intelligence); Wiener over-learning problem; feedforward neural networks; inverse problem; memorization learning; necessary condition; sufficient condition; training sets; Computer science; Feedforward neural networks; Hilbert space; Inverse problems; Kernel; Multi-layer neural network; Neural networks; Sufficient conditions; Training data; Vectors;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374224
