Dropout is Drop-Dead Easy to Implement

It’s about 1 line of python…

Figure from the original paper describing dropout.

Introduction

We’ve all heard of dropout. Historically it’s one of the most famous ways of regularizing a neural network, though nowadays it’s fallen somewhat out of favor and has been replaced by batch normalization. Regardless, it’s a cool technique and very simple to implement in PyTorch. In this quick blog post…