AN ADAPTIVE HYPERSPHERE DECISION BOUNDARY.
SYLVANIA ELECTRONIC SYSTEMS WALTHAM MASS APPLIED RESEARCH LAB
Pagination or Media Count:
It was previously shown that the hypersphere decision boundary optimally partitions an n-dimensional sample space when the underlying category probability distributions are of certain types, including spherically symmetric normal and Pearson Types II and VII. Such a model could apply to pattern recognition or to detection of band-limited white stochastic signals. The hypersphere partition arises for distributions differing in their means and in their variances. This paper examines the problem of adaptation, and treats the following two topics non-supervised adaptation to the optimum hypersphere for normal distributions supervised estimation of the Pearson shape parameter m, thereby supplementing the partial treatment of estimation in an earlier hypersphere paper. Author
- Statistics and Probability