An Algebraic Approach to Conditioning in Probability with Applications to the Combination of Evidence Problem.

Abstract

This presentation considers a fundamental problem touching upon four major disciplines: probability theory, boolean algebra and logic, ring theory, and the modeling of natural and formal language in expert systems. Specifically, this lecture treats the problem of annexing a conditional event operator to boolean algebra-as originally proposed by Boole and long neglected in the standard semantically oriented literature-which is compatible with all conditional probability evaluations, and which allows for the development of a full calculus of extended boolean operations and relations. In this lecture a new procedure is presented for dealing with the combination of such implicational statements, derived from a minimum of assumptions. This yields a sound and complete logic-conditional probability logic of propositions-which has connections with Koopman's qualitative conditional probability and possesses the algebraic structure of semi-boolean algebra. Extensive applications to use in expert systems are also exhibited, showing how this new approach to conditioning can be used in design of knowledge-based systems and the treatment of uncertainty factors.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 1991
Accession Number
ADA240879

Entities

People

  • I. R. Goodman

Tags

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Boolean Algebra
  • Calculus
  • Computer Languages
  • Expert Systems
  • Formal Languages
  • Fuzzy Sets
  • Knowledge Based Systems
  • Language
  • Logic
  • Mathematics
  • Probability
  • Uncertainty

Readers

  • Graph Algorithms and Convex Optimization.
  • Mathematical Modeling and Probability Theory.
  • Theoretical Analysis.