CONSTRUCTION OF MINIMAL N-2-N ENCODERS FOR ANY N

Citation
Ds. Phatak et al., CONSTRUCTION OF MINIMAL N-2-N ENCODERS FOR ANY N, Neural computation, 5(5), 1993, pp. 783-794
Citations number
3
Categorie Soggetti
Computer Sciences","Computer Applications & Cybernetics",Neurosciences
Journal title
ISSN journal
08997667
Volume
5
Issue
5
Year of publication
1993
Pages
783 - 794
Database
ISI
SICI code
0899-7667(1993)5:5<783:COMNEF>2.0.ZU;2-L
Abstract
The encoding problem (Rumelhart and McClelland 1986) is an important c anonical problem. It has been widely used as a benchmark. Here, we hav e analytically derived minimal-sized nets necessary and sufficient to solve encoding problems of arbitrary size. The proofs are constructive : we construct n-2-n encoders and show that two hidden units are also necessary for n > 2. Moreover, the geometric approach employed is gene ral and has much wider applications. For example, this method has also helped us derive lower bounds on the redundancy necessary for achievi ng complete fault tolerance (Phatak and Koren 1992a,b).