Use of One-Point Coverage Representations, Product Space Conditional Event Algebra, and Second-Order Probability Theory for Constructing and Using Probability-Compatible Inference Rules in Data-Fusion Problems

摘要

This paper covers issues relating to the establishment of a sound and conditional probability-compatible rationale for generating linguistic- based inference rules concerning a population. By extending previous preliminary results, the authors detail, in a fully rigorous manner and within the confines of traditional probability theory, that a comprehensive technique can be derived that converts linguistic-based conditional information, couched only in fuzzy-logic terms, into naturally corresponding conditional probabilities. In turn, they demonstrate how such typically underconstrained conditional probabilities can be combined for suitable conclusions and decision making, via a new use of second-order probability logic. This research is part of the ongoing SSC San Diego In-house Laboratory Independent Research FY 01 project CRANOF (a Complexity-Reducing Algorithm for Near-Optimal Fusion).