Figure 3.

Learning rates of models trained on the different sized training sets. Each model is trained on 25, 50, 75, and 100 percent of relevent data sets. AUROCs were binned into four equal-interval by the number of training cases.