@inproceedings{88033a4d28bd47b3a67f9c83b13a0cb9,
title = "Pattern classification based on local learning",
abstract = "Local learning methods approximate a target function (a posteriori probability) by partitioning the input space into a set of local regions, and modeling a simple input-output relationship in each one. In order for local learning to be effective for pattern classification in high dimensional settings, regions must be chosen judiciously to minimize bias. This paper presents a novel region partitioning criterion that attempts to minimize bias by capturing differential relevance in input variables in an efficient way. The efficacy of the method is validated using a variety of real and simulated data.",
author = "Jing Peng and Bir Bhanu",
year = "1998",
month = jan,
day = "1",
language = "English",
isbn = "3540648585",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "882--889",
editor = "Adnan Amin and Dov Dori and Pavel Pudil and Herbert Freeman",
booktitle = "Advances in Pattern Recognition - Joint IAPR International Workshops SSPR 1998 and SPR 1998, Proceedings",
note = "7th Joint IAPR International Workshop on Structural and Syntactic Pattern Recognition, SSPR 1998 and 2nd International Workshop on Statistical Techniques in Pattern Recognition, SPR 1998 ; Conference date: 11-08-1998 Through 13-08-1998",
}