Split-merge incremental learning (SMILE) of mixture models

Authors:
Konstantinos Blekas;Isaac E. Lagaris
Affiliations:
Department of Computer Science, University of Ioannina, Ioannina, Greece;Department of Computer Science, University of Ioannina, Ioannina, Greece
Venue:
ICANN'07 Proceedings of the 17th international conference on Artificial neural networks
Year:
2007

Citing 7
Cited 3

Introduction to statistical pattern recognition (2nd ed.)

Introduction to statistical pattern recognition (2nd ed.)
Unsupervised Learning of Finite Mixture Models

IEEE Transactions on Pattern Analysis and Machine Intelligence
Pattern Recognition and Neural Networks

Pattern Recognition and Neural Networks
A Greedy EM Algorithm for Gaussian Mixture Learning

Neural Processing Letters
Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Pattern Recognition and Machine Learning (Information Science and Statistics)

Pattern Recognition and Machine Learning (Information Science and Statistics)
SMEM Algorithm for Mixture Models

Neural Computation

Incremental semi-supervised clustering in a data stream with a flock of agents

CEC'09 Proceedings of the Eleventh conference on Congress on Evolutionary Computation
Geo-temporal structuring of a personal image database with two-level variational-bayes mixture estimation

AMR'08 Proceedings of the 6th international conference on Adaptive Multimedia Retrieval: identifying, Summarizing, and Recommending Image and Music
Fast intelligent watermarking of heterogeneous image streams through mixture modeling of PSO populations

Applied Soft Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this article we present an incremental method for building a mixture model. Given the desired number of clusters K ≥ 2, we start with a two-component mixture and we optimize the likelihood by repeatedly applying a Split-Merge operation. When an optimum is obtained, we add a new component to the model by splitting in two, a properly chosen cluster. This goes on until the number of components reaches a preset limiting value. We have performed numerical experiments on several data-sets and report a performance comparison with other rival methods.