FIGURE 2.

Feature selection scheme. From left to right, training data from partitions 1-5 are fed to a GA. The GA yields a solution for each partition (“Selected features”) and finally, the shared features between those solutions are collected (“Common features”).