Feature Selection for Local Learning Based Clustering

Authors:
Hong Zeng;Yiu-Ming Cheung
Affiliations:
Department of Computer Science, Hong Kong Baptist University, Hong Kong SAR, China;Department of Computer Science, Hong Kong Baptist University, Hong Kong SAR, China
Venue:
PAKDD '09 Proceedings of the 13th Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining
Year:
2009

Citing 6
Cited 2

Feature Selection for Clustering - A Filter Solution

ICDM '02 Proceedings of the 2002 IEEE International Conference on Data Mining
Multiclass Spectral Clustering

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Feature Selection for Unsupervised Learning

The Journal of Machine Learning Research
Feature Selection for Unsupervised and Supervised Inference: The Emergence of Sparsity in a Weight-Based Approach

The Journal of Machine Learning Research
Feature Selection for Clustering on High Dimensional Data

PRICAI '08 Proceedings of the 10th Pacific Rim International Conference on Artificial Intelligence: Trends in Artificial Intelligence
Local Kernel Regression Score for Selecting Features of High-Dimensional Data

IEEE Transactions on Knowledge and Data Engineering

Kernel Learning for Local Learning Based Clustering

ICANN '09 Proceedings of the 19th International Conference on Artificial Neural Networks: Part I
Feature selection for transfer learning

ECML PKDD'11 Proceedings of the 2011 European conference on Machine learning and knowledge discovery in databases - Volume Part III

Quantified Score

Hi-index	0.00

Visualization

Abstract

For most clustering algorithms, their performance will strongly depend on the data representation. In this paper, we attempt to obtain better data representations through feature selection, particularly for the Local Learning based Clustering (LLC) [1]. We assign a weight to each feature, and incorporate it into the built-in regularization of LLC algorithm to take into account of the relevance of each feature for the clustering. Accordingly, the weights are estimated iteratively with the clustering. We show that the resulting weighted regularization with an additional constraint on the weights is equivalent to a known sparse-promoting penalty, thus the weights for irrelevant features can be driven towards zero. Experiments on several benchmark datasets demonstrate the effectiveness of the proposed method.