REDUCED-SIZE NEURAL NETWORKS THROUGH SINGULAR-VALUE DECOMPOSITION AND SUBSET-SELECTION

被引：18

作者：

KANJILAL, PP

DEY, PK

BANERJEE, DN

机构：

[1] Dept. of Electronics & ECE, IIT

来源：

ELECTRONICS LETTERS | 1993年 / 29卷 / 17期

关键词：

NEURAL NETWORKS; MATRIX ALGEBRA;

D O I：

10.1049/el:19931010

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

Singular value decomposition (SVD) can lead to compaction of information, and along with QR with column pivoting (QRcp) factorisation can be used for subset selection. The Letter proposes an application of SVD and subset selection for optimising the size of feedforward neural networks; the Mackey Glass (MG) series is used as an example.

引用

页码：1516 / 1518

页数：3

共 5 条

[1] AN INFORMATION CRITERION FOR OPTIMAL NEURAL NETWORK SELECTION
FOGEL, DB
[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 1991, 2 (05): : 490 - 497
[2] GOLUB GH, 1989, MATRIX COMPUTATIONS
[3] OSCILLATION AND CHAOS IN PHYSIOLOGICAL CONTROL-SYSTEMS
MACKEY, MC
GLASS, L
[J]. SCIENCE, 1977, 197 (4300) : 287 - 288
[4] A Tree-Structured Algorithm for Reducing Computation in Networks with Separable Basis
Sanger, Terence D.
[J]. NEURAL COMPUTATION, 1991, 3 (01) : 67 - 78
[5] Tenorio M F, 1990, IEEE Trans Neural Netw, V1, P100, DOI 10.1109/72.80209

← 1 →