Learning contiguity with layered neural networks

Sara A. Solla*

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review

Abstract

The problem of inductive inference refers to extracting general rules, learning concepts from a training set consisting of some fraction of the total number of possible examples of the concept. This work explores the ability of a highly connected, layered network of simple analog processing units to perform such task. The problem as posed is ill-defined, since there are in general many concepts consistent with a given training set. Precise definitions of learning and generalization are given and incorporated into a measure of the efficiency with which a network learns from examples. Specific network performance is analyzed for the contiguity problem, for which a formal solution based on the idea of edge detection demonstrates the second order of the predicate. A simple learning algorithm applied to a network of second order units finds such solution without difficulty, since the adaptive part of the network is thus reduced to a one-layer perceptron.

Original languageEnglish (US)
Number of pages1
JournalNeural Networks
Volume1
Issue number1 SUPPL
DOIs
StatePublished - Jan 1 1988
EventInternational Neural Network Society 1988 First Annual Meeting - Boston, MA, USA
Duration: Sep 6 1988Sep 10 1988

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Learning contiguity with layered neural networks'. Together they form a unique fingerprint.

Cite this