Unbiased Feature Selection in Learning Random Forests for High-Dimensional Data

Figure 3

Box plots: the test accuracy of the nine Caltech subdatasets.