Modularized Implementation

https://youtu.be/d14TUNcbn1k?list=PL3FW7Lu3i5JvHM8ljYj-zLfQRF3EO8sYv&t=3019

We have a forwards and a backwards API with each function computing the respective pass

In assignment 1 you’ll get practice with this computational graph way of thinking where you’ll be writing your SVM and softmax losses and taking the gradients of these.

First step represent it as a computational graph and when it’s time to do your backward pass take gradientw respect to each variables that you’ve represented in your computational graph and use the chain rule to link them together