Feature Weighted Rival Penalized EM for Gaussian Mixture Clustering: Automatic Feature and Model Selections in a Single Paradigm

Authors:
Yiu-Ming Cheung;Hong Zeng
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:
Computational Intelligence and Security
Year:
2007

Citing 8
Cited 0

Visualization and interactive feature selection for unsupervised data

Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining
Knowledge Acquisition Via Incremental Conceptual Clustering

Machine Learning
Feature Selection and Incremental Learning of Probabilistic Concept Hierarchies

ICML '00 Proceedings of the Seventeenth International Conference on Machine Learning
Simultaneous Feature Selection and Clustering Using Mixture Models

IEEE Transactions on Pattern Analysis and Machine Intelligence
A Rival Penalized EM Algorithm towards Maximizing Weighted Likelihood for Density Mixture Clustering with Automatic Model Selection

ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 4 - Volume 04
Automated Variable Weighting in k-Means Type Clustering

IEEE Transactions on Pattern Analysis and Machine Intelligence
Maximum Weighted Likelihood via Rival Penalized EM for Density Mixture Clustering with Automatic Model Selection

IEEE Transactions on Knowledge and Data Engineering
Bayesian Feature and Model Selection for Gaussian Mixture Models

IEEE Transactions on Pattern Analysis and Machine Intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

Recently, the Rival Penalized Expectation-Maximization (RPEM) algorithm (Cheung 2004 & 2005) has demonstrated its outstanding capability to perform the model selection automatically in the context of density mixture models. Nevertheless, the RPEM is unable to exclude the irrelevant variables (also called features) from the clustering process, which may degrade the algorithm's performance. In this paper, we adopt the concept of feature salience (Law et al. 2004) as the feature weight to measure the relevance of features to the cluster structure in the subspace, and integrate it into the RPEM algorithm. The proposed algorithm identifies the irrelevant features and estimates the number of clusters automatically and simultaneously in a single learning paradigm. Experiments show the efficacy of the proposed algorithm on both synthetic and benchmark real data sets.