Figure 3
From: To aggregate or not to aggregate high-dimensional classifiers
Boxplot of cross-validation errors for three real data sets. Misclassification rates are obtained from 1000 times repeating 10 fold double cross-validations. The ratios of feature to samples in training sets are 38/7129 (Leukemia), 39/500(Gaucher) and 200/412 (Grape). The most stable case is from grape data, and the ratio of feature to sample is the lowest among all three data sets.