Example of learning curves obtained during training. The black and gray curves indicate the performance (in terms of loss) of a model on both the training (black) and test (gray) set. The best model could either be the model that scored the lowest loss on the test set (red dashed line), or the average lowest loss on both the training and test sets (blue dashed line). Notice that after about 50 epochs the test loss starts to increase, indicating a possible overfitting of the training data.