$($c$)$ Denote with CNN$1$ the network in the code, and with CNN$2$ another alternative

CNN$.$ Both network are trained on the training data and assessed on a separate

validation set. Figure $6$ below reports the learning curves corresponding to

the loss values of the training and validation procedure for both networks.

Figure $6$

The network CNN$2$ has the same architecture, learning rate, and batch size as

CNN$1,$ but it is implemented with a major change to the training process. Com$-$

ment on the figure and explain which change has most likely been implemented

on the basis of the context and the figure.