Repairing Learned Knowledge Using Experience

Abstract

Explanation-based learning occurs when something useful is retained from an explanation exercise, usually an explanation of how some particular problem can be solved. If explanation is based on sound theory, then the learning process speeds up future problem solving, but the scope of the learning-augmented theory remains unchanged. In contrast, if explanation is based on fragmentary and often defective experience, explanation can be a guide to when that experience can be deployed. Thus one kind of explanation provides speed up; another kind of explanation provides new knowledge. Experience is not sound theory, however, and wrong things may be learned accidentally, as subsequent failures will likely demonstrate. In this paper, we describe ways to isolate the facts that cause failures, ways to explain why those facts cause problems. and ways to repair learning mistakes. In particular, our program learns to distinguish pails from cups after learned knowledge about cups leads a recognition program to think pails are cups.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 1990
Accession Number
ADA228711

Entities

People

  • Patrick Winston
  • Satyajit Rao

Organizations

  • Massachusetts Institute of Technology

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Databases
  • Digestive System Processes
  • Information Systems
  • Learning
  • Military Research
  • Porcelain
  • Reasoning
  • Recognition

Readers

  • Artificial Intelligence
  • Theoretical Analysis.