Distributions of the occurrence of individual features in the top10 feature groups for each given size of the feature set from 1 (top left panel) to 16 (bottom right panel) features.

<p>If a single feature is used, feature 4 (CV of spike width) is best for data set 1 (A) and feature 14 (minimum spike slope) for data set 2 (B), but when using 2 or more features, combinations involving other features work best. The highly successful feature sets between 5 to 10 features (second row) show some commonalities in that for data set 1 (A) features 8 (std of noise level) and 13 (maximum spike slope) are always included and for data set 2 (B) features 1 (mean spike height), 3 (CV of spike height), 8 (std of noise level) and 14 (minimum spike slope) play a dominant role. It is noteworthy that observations reported here do not seem to generalize well between the two data sets used in this study.