TRAINING FEEDFORWARD NETWORKS WITH THE MARQUARDT ALGORITHM

被引：5495

作者：

HAGAN, MT

MENHAJ, MB

机构：

来源：

IEEE TRANSACTIONS ON NEURAL NETWORKS | 1994年 / 5卷 / 06期

关键词：

D O I：

10.1109/72.329697

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

The Marquardt algorithm for nonlinear least squares is presented and is incorporated into the backpropagation algorithm for training feedforward neural networks. The algorithm is tested on several function approximation problems, and is compared with a conjugate gradient algorithm and a variable learning rate algorithm. It is found that the Marquardt algorithm is much more efficient than either of the-other techniques when the network contains no more than a few hundred weights.

引用

页码：989 / 993

页数：5

共 16 条

[1] [Anonymous], 1988, INT C NEUR INF PROC
[2] OPTIMIZATION FOR TRAINING NEURAL NETS
BARNARD, E
[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 1992, 3 (02): : 232 - 240
[3] 1ST-ORDER AND 2ND-ORDER METHODS FOR LEARNING - BETWEEN STEEPEST DESCENT AND NEWTON METHOD
BATTITI, R
[J]. NEURAL COMPUTATION, 1992, 4 (02) : 141 - 166
[4] CONJUGATE-GRADIENT ALGORITHM FOR EFFICIENT TRAINING OF ARTIFICIAL NEURAL NETWORKS
CHARALAMBOUS, C
[J]. IEE PROCEEDINGS-G CIRCUITS DEVICES AND SYSTEMS, 1992, 139 (03): : 301 - 310
[5] INCREASED RATES OF CONVERGENCE THROUGH LEARNING RATE ADAPTATION
JACOBS, RA
[J]. NEURAL NETWORKS, 1988, 1 (04) : 295 - 307
[6] AN ADAPTIVE LEAST-SQUARES ALGORITHM FOR THE EFFICIENT TRAINING OF ARTIFICIAL NEURAL NETWORKS
KOLLIAS, S
ANASTASSIOU, D
[J]. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, 1989, 36 (08): : 1092 - 1101
[7] ON THE LIMITED MEMORY BFGS METHOD FOR LARGE-SCALE OPTIMIZATION
LIU, DC
NOCEDAL, J
[J]. MATHEMATICAL PROGRAMMING, 1989, 45 (03) : 503 - 528
[8] AN ALGORITHM FOR LEAST-SQUARES ESTIMATION OF NONLINEAR PARAMETERS
MARQUARDT, DW
[J]. JOURNAL OF THE SOCIETY FOR INDUSTRIAL AND APPLIED MATHEMATICS, 1963, 11 (02): : 431 - 441
[9] NGUYEN D, 1990, P INT JOINT C NEUR N, V3, P21, DOI DOI 10.1109/IJCNN.1990.137819
[10] PUSKORIUS GV, 1991, JUL P IJCNN, V1, P771

← 1 2 →