Of course, looking at the graph above it does appear that we're getting near saturation. Suppose, however, that we redo the graph with the training set size plotted logarithmically:It seems clear that the graph is still going up toward the end. This suggests that if we used vastly more ...