Uniqueness of the weights for minimal feedforward nets with a given input-output map

Héctor J. Sussmann

Research output: Contribution to journalArticlepeer-review

193 Scopus citations

Abstract

Abstract: We show that, for feedforward nets with a single hidden layer, a single output node, and a "transfer function" Tanh s, the net is uniquely determined by its input-output map, up to an obvious finite group of symmetries (permutations of the hidden nodes, and changing the sign of all the weights associated to a particular hidden node), provided that the net is irreducible (i.e., that there does not exist an inner node that makes a zero contribution to the output, and there is no pair of hidden nodes that could be collapsed to a single node without altering the inputoutput map).

Original languageEnglish (US)
Pages (from-to)589-593
Number of pages5
JournalNeural Networks
Volume5
Issue number4
DOIs
StatePublished - 1992

All Science Journal Classification (ASJC) codes

  • Cognitive Neuroscience
  • Artificial Intelligence

Keywords

  • Feedforward nets
  • Symmetries
  • Uniqueness

Fingerprint

Dive into the research topics of 'Uniqueness of the weights for minimal feedforward nets with a given input-output map'. Together they form a unique fingerprint.

Cite this