Unsupervised Ensembles Techniques for Visualization

In this paper we introduce two unsupervised techniques for visualization purposes based on the use of ensemble methods. The unsupervised techniques which are often quite sensitive to the presence of outliers are combined with the ensemble approaches in order to overcome the influence of outliers. The first technique is based on the use of Principal Component Analysis and the second one is known for its topology preserving characteristics and is based on the combination of the Scale Invariant Map and Maximum Likelihood Hebbian learning. In order to show the advantage of these novel ensemble-based techniques the results of some experiments carried out on artificial and real data sets are included.