Title :
Analysis of statistical training method
Author :
Liang, Xun ; XIA, Shaowei
Author_Institution :
Dept. of Autom., Tsinghua Univ., Beijing, China
Abstract :
This paper analyzes the statistical method of training multilayer perceptrons (MLPs) with binary patterns. First, several MLPs are trained until they are trapped in local minima; then they are fabricated into the demanding one. The average number of the MLPs which may be used in primary training is estimated. Two examples are given and techniques in fabricating are discussed.
Keywords :
learning (artificial intelligence); multilayer perceptrons; probability; statistical analysis; binary patterns; local minima; multilayer perceptrons; probability; statistical training; Automation; Laboratories; Management information systems; Multilayer perceptrons; Neurons; Pattern analysis; Statistical analysis;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.716970
