NONSUPERVISED ADAPTIVE DETECTION FOR MULTIVARIATE NORMAL DISTRIBUTIONS.
Final rept., Feb 65-Apr 66,
SYLVANIA ELECTRONIC SYSTEMS-EAST WALTHAM MASS
Pagination or Media Count:
Nonsupervised adaptive detection for categories described statistically by multivariate normal distributions is approached as a problem in multi-parameter estimation for a multi-modal distribution. Nonsupervised learning consists in estimating the component probability distributions of a mixture of distributions, given a sequence of samples known only to have been drawn from the over-all mixture. This report considers the two-category case involving general unequal covariance matrices and the multiple-category case for spherically symmetric distributions. The techniques provided are applicable to problems in statistical classification, pattern recognition, and signal detection. Author
- Statistics and Probability