HW5 Q1.4: Training a Model


When we need to train a binary classifier, should we be building everything from scratch (modifying the classes to create our own MLPs and then using functions like those from hw4 to train the model) or can we use gluon and d2l to train the model (e.g. based on the Softmax Regression in Gluon notebook from 2/5)?



You can do either – using the dense layers implemented in Gluon is probably easiest.