A Neural Network Training Method Based On Error Variety Rate

This paper presents a neural network training method used for function fitting. The traditional performance index for neural network training based solely on sum of squared error, it is possible to lead to the inferior fitting results, and as a consequence, this paper defines a new performance index...

Full description

Saved in:
Bibliographic Details
Published in2021 IEEE Sustainable Power and Energy Conference (iSPEC) pp. 3960 - 3963
Main Authors Enze, Shao, Zihan, Wang, Can, Wang, Xiao, Xu, Xianbo, Du, Chunlin, Zhong, Lei, Zou, Zhengyong, Wu, Chao, Fang
Format Conference Proceeding
LanguageEnglish
Published IEEE 23.12.2021
Subjects
Conferences
Fitting
Function Fitting
Indexes
Multi-Objective Optimization
Neural Network
Neural networks
Performance analysis
Performance Index
Simulation
Training
Online AccessGet full text

Cover

More Information
Summary:This paper presents a neural network training method used for function fitting. The traditional performance index for neural network training based solely on sum of squared error, it is possible to lead to the inferior fitting results, and as a consequence, this paper defines a new performance index based on the rate of variation between network output and desired output. This paper shows the method of neural network training which is based on these two performance indexes. The simulation result shows the feasibility.
DOI:10.1109/iSPEC53008.2021.9735544