Average M_c with SEM as error bars for the last-layer, every-layer, and scratch models. The M_c of the every-layer models are slightly higher than those of the scratch models. The last-layer models show higher M_c for small training sets, and vice versa for large training sets. We add a small offset on the x-axis to the scratch models in order to prevent its error bars from lying on top of the other error bars.