Clustering in a Boosting Framework
Amir Saffari, Horst Bischof · 2008
Abstract In this paper we present a novel approach for creating partitions of data space using simple clustering algorithms in a boosting framework. A general boosting algorithm for clustering tasks is proposed, and solutions for directly optimizing two loss functions according to this framework are obtained. Experimental results show how the performance of relatively simple and computationally efficient base clustering algorithms could be boosted using the proposed algorithm. 1