Subfigure 6(a): Colon cancer; Subfigure 6(b): Lymphoma; Subfigure 6(c): Tumor vs. metastases. The accumulated number of extractions (gene multiplicity) in 50 replicated runs is compared to the f
weight ranking function for features selected by the E-RFE model trained on the whole data set at n* features. The groups of isolated points in the upper part of the panels indicate variables in the optimal feature sets for the replicated and not extracted in the global model. The variables labeled by capital Latin letter are those extracted more than 25 times in the replicates. The dotted line marks the 25 extractions threshold.