Linear neural network training algorithms for real-world benchmark problems

被引：0
|
作者：
Goulianas, K [1 ]
Adamopoulos, M
Katsavounis, S
Fragakis, C
Tsouros, CC
机构：
[1] Technol Educ Inst Thessaloniki, Dept Informat, Thessaloniki, Greece
[2] Univ Macedonia, Dept Informat, Thessaloniki, Greece
[3] Aristotle Univ Thessaloniki, Fac Engn, Thessaloniki, Greece
来源：
INTERNATIONAL JOURNAL OF COMPUTER MATHEMATICS | 2002年 / 79卷 / 11期
关键词：
neural nets; training algorithms; iterative methods;
D O I：
10.1080/00207160213945
中图分类号：
O29 [应用数学];
学科分类号：
070104 ;
摘要：
This paper describes the Adaptive Steepest Descent (ASD) and Optimal Fletcher-Reeves (OFR) algorithms for linear neural network training. The algorithms are applied to well-known pattern classification and function approximation problems, belonging to benchmark collection Proben1. The paper discusses the convergence behavior and performance of the ASD and OFR training algorithms by computer simulations and compares the results with those produced by linear-RPROP method.
引用
收藏
页码：1149 / 1167
页数：19
相关论文
共 50 条
12345