DETERMINING THE STATISTICAL CHARACTERISTICS OF DISCERNIBLE IMAGES IN THE SELF-INSTRUCTION MODE,
Abstract
The need to employ the learning process (also termed self-learning or learning without a teacher) arises in many important practical cases where a priori information on patterns is insufficiently complete or insufficiently reliable and when, for one reason or another, the customary teaching procedure cannot be organized. In such cases the division of pattern realizations into classes is based on intuitively introduced measures of compactness. A more rigorous approach, however, is that of estimating the statistical characteristics of the recognized patterns with respect to the totality of the incoming realizations and then classifying them in accordance with these estimates on the basis of Bayes decision rules. Such an approach makes it possible to optimize the division and to use many of the findings obtained with the aid of the theory of statistical decisions in a broad class of problems involving limited a priori information.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 04, 1968
- Accession Number
- AD0683575
Entities
People
- A. V. Milenkii
Organizations
- National Air and Space Intelligence Center