Locally centralizing samples for nearest neighbors

Authors:
Guihua Wen;Si Wen;Jun Wen;Lijun Jiang
Affiliations:
South China University of Technology, Guangzhou, China;South China University of Technology, Guangzhou, China;Hubei Institute for Nationalities, Ensi, China;South China University of Technology, Guangzhou, China
Venue:
PRICAI'10 Proceedings of the 11th Pacific Rim international conference on Trends in artificial intelligence
Year:
2010

Citing 10
Cited 0

2D spiral pattern recognition with possibilistic measures

Pattern Recognition Letters
A Bootstrap Technique for Nearest Neighbor Classifier Design

IEEE Transactions on Pattern Analysis and Machine Intelligence
Automatic Textual Document Categorization Based on Generalized Instance Sets and a Metamodel

IEEE Transactions on Pattern Analysis and Machine Intelligence
Adaptive Quasiconformal Kernel Nearest Neighbor Classification

IEEE Transactions on Pattern Analysis and Machine Intelligence
Nearest Neighbors by Neighborhood Counting

IEEE Transactions on Pattern Analysis and Machine Intelligence
A local mean-based nonparametric classifier

Pattern Recognition Letters
Using locally estimated geodesic distance to optimize neighborhood graph for isometric data embedding

Pattern Recognition
Local relative transformation with application to isometric embedding

Pattern Recognition Letters
Tailored Aggregation for Classification

IEEE Transactions on Pattern Analysis and Machine Intelligence
The Nearest Neighbor Algorithm of Local Probability Centers

IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

The k nearest neighbors classifier is simple and often results in good performance in problems. However, it can not work well on noisy and high dimensional data, as the structure composed of selected nearest neighbors on these data is easily deformed and perceptually unstable. This paper presents a locally centralizing samples approach with kernel techniques to preprocess the data. It creates a new sample for each original sample through its neighborhood and then replace it to be candidate for nearest neighbors. This approach can be justified by gestalt psychology and applied to provide better quality data for classifiers, even if the original data is noisy and high dimensional. The conducted experiments on challenging benchmark data sets validate the proposed approach.