DEVELOPMENT AND EVALUATION OF PROCEDURES FOR QUANTIZING MULTIVARIATE DISTRIBUTIONS,
Abstract
The approximation generated by a specific but arbitrary partition and a specific but arbitrary set of representation points is examined. A statistical measure of error is investigated. The size of the error is measured by some moment of the random variable. Such a measure involves both the geometrical properties of the partition ing in the k-dimensional space, and the probabil ity density. The approximation of the random variable is accomplished by selecting the representation points at random, independently of one another, and according to some given density function and then constructing the parti tion. In Chapter 2 we shall examine the behavior of the lower bound of the average normalized error for fixed variables is examined. The behavior of the minimum error when there is a large number of representation points is demonstrated. This extension takes into account the entropy rate of the discrete quantizing sequence. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 10, 1963
- Accession Number
- AD0426677
Entities
People
- Paul Zador
Organizations
- Stanford University