Component vs. All Learning - Summary

It appears, from our experiments, that learning on dense components and testing on all components wins over the traditional learning on all, and testing on all. Here is the summary table.