On the efficiency of swap-based clustering

Authors:
Pasi Fränti;Olli Virmajoki
Affiliations:
Department of Computer Science, University of Joensuu, Finland;Department of Computer Science, University of Joensuu, Finland
Venue:
ICANNGA'09 Proceedings of the 9th international conference on Adaptive and natural computing algorithms
Year:
2009

Citing 6
Cited 0

The LBG-U Method for Vector Quantization – an Improvement over LBGInspired from Neural Networks

Neural Processing Letters
A local search approximation algorithm for k-means clustering

Computational Geometry: Theory and Applications - Special issue on the 18th annual symposium on computational geometry—SoCG2002
Iterative shrinking method for clustering problems

Pattern Recognition
Improvement of the k-means clustering filtering algorithm

Pattern Recognition
Clustering by competitive agglomeration

Pattern Recognition
A fast exact GLA based on code vector activity detection

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Random swap-based clustering is very simple to implement and guaranteed to find the correct clustering if iterated long enough. However, its quadratic dependency on the number of clusters can be too slow in case of some data sets. Deterministic selection of the swapped prototype can speed-up the algorithm but only if the swap can be performed fast enough. In this work, we introduce an efficient implementation of the swap-based heuristic and compare its time-distortion efficiency against random and deterministic variants of the swap-based clustering, and repeated k-means.