PLAIN BACKPROPAGATION AND ADVANCED OPTIMIZATION ALGORITHMS - A COMPARATIVE-STUDY

被引：4

作者：

DEGROOT, C

WURTZ, D

机构：

[1] SWISS FED INST TECHNOL,INTERDISZIPLINARES PROJEKTZENTRUM SUPERCOMP,CH-8092 ZURICH,SWITZERLAND

[2] ETH HONGGERBERG,INST THEORET PHYS,CH-8092 ZURICH,SWITZERLAND

来源：

NEUROCOMPUTING | 1994年 / 6卷 / 02期

关键词：

BACKPROPAGATION; FEEDFORWARD NEURAL NETWORKS; OPTIMIZATION;

D O I：

10.1016/0925-2312(94)90052-3

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We present a comparison of 'plain backpropagation' to other standard optimization algorithms. These algorithms include direct search as well as first and second order methods. All algorithms share the property of easy accessability. We chose a feedforward architecture for a network to perform function approximation. The results show a clear advantage of higher order methods with respect to quality of solution as well as execution times.

引用

页码：153 / 161

页数：9

共 17 条

[1] OPTIMIZATION FOR TRAINING NEURAL NETS [J].