Parallel Levenberg-Marquardt Algorithm Without Error Backpropagation

被引：10

作者：

Bilski, Jaroslaw ^{[1
]}

Wilamowski, Bogdan M. ^{[2
]}

机构：

[1] Czestochowa Tech Univ, Inst Computat Intelligence, Czestochowa, Poland

[2] Auburn Univ, Auburn, AL 36849 USA

来源：

ARTIFICIAL INTELLIGENCE AND SOFT COMPUTING, ICAISC 2017, PT I | 2017年 / 10245卷

关键词：

Forward-only computation; Neural network training; Parallel architectures; REALIZATION; NETWORKS;

D O I：

10.1007/978-3-319-59063-9_3

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This paper presents a new parallel architecture of the Levenberg-Marquardt (LM) algorithm for training fully connected feedforward neural networks, which will also work for MLP but some cells will stay empty. This approach is based on a very interesting idea of learning neural networks without error backpropagation. The presented architecture is based on completely new parallel structures to significantly reduce a very high computational load of the LM algorithm. A full explanation of parallel three-dimensional neural network learning structures is provided.

引用

页码：25 / 39

页数：15

共 34 条

[1]

[Anonymous], 2017, DESIGN INTERPRETABLE

[2]

[Anonymous], 1988, Parallel distributed processing

[3]

[Anonymous], P CONN MOD SUMM SCH

[4]

[Anonymous], 1993, IEEE INT C NEUR NETW

[5]

Bilski J, 2004, LECT NOTES ARTIF INT, V3070, P151

[6]

Bilski J, 2004, LECT NOTES ARTIF INT, V3070, P158

[7]

Bilski J., 2004, INT J APPL MATH COMP, V3070, P158

[8]

Bilski J., 2003, ADV SOFT COMPUTING, V19, P3

[9]

Bilski J., 2013, PARALLEL STRUCTURES

[10]

Bilski J, 2008, LECT NOTES ARTIF INT, V5097, P11, DOI 10.1007/978-3-540-69731-2_2

← 1 2 3 4 →