Fig. 2From: Automatic discovery of 100-miRNA signature for cancer classification using ensemble feature selectionHeatmap with the frequency of the overall top 100 most frequent features, divided by classifier. Features are sorted from overall most to least frequent, from left to right, using information from the whole ensemble. For example, the most frequent is mir-10b, that is considered important by all classifiers. Color intensity is computed using information from instances of the same classifier, only. This shows the different importance that different classifiers assign to each featureBack to article page