Fig. 13

High- and low-performing features for the IM data. Ellipses representing clusters for features in classification accuracy against SI plots for results using the IM data set. The SI distance definition is modified Mahalanobis. The ellipses are centered around the means of the feature clusters and constructed according to their covariance matrices. Every inset includes the features with the top five and bottom five average classification accuracies for the classifier stated in the plot. The ellipses are coded by red and blue color for low and high average classification accuracy, respectively. Classifiers were used in the conventional “single” topology, apart from LDA, which was used in “single” and OVO