This is a fundamental research project in the area of machine learning. Accordingly, the goals are of methodological, theoretical and algorithmic nature. More specifically, the project aims at:
- developing cluster combination techniques, in which partitions, clusters and even samples can assume different weights in the combination strategies.
- integrating common and state-of-the-art clustering algorithms in the combination process
- creating a theoretical framework for the analysis of clustering combination techniques and cluster validity
- extending combination techniques used in supervised and unsupervised learning to the semi-supervised scenario.
- applying the developed methodologies to challenging learning problems (both unsupervised and semi-supervised), such as document classification, web page classification, and gene-expression data analysis.