The losses of the training process before and after the optimization of loss function.

A blue and an orange dot lines represent losses of training and validation set before optimization, whereas a yellow and a light blue lines represent losses of training and validation set after optimization. A gray line is the line when loss = 0.05.