Local polynomial models for classification
Ron Meir, Robert R. Snapp · 21st IEEE Convention of the Electrical and Electronic Engineers in Israel. Proceedings (Cat. No.00EX377) · 2002
We consider the problem of pattern classification based on empirical data. In particular, we focus on local approaches (Fan and Gijbels 1996), based on information obtained from the neighborhood of each point whose classification is desired, similar to the nearest neighbor approach. The problem is formulated within a local maximum-likelihood approach, and performance bounds as well as some preliminary empirical simulation results are presented.