Feedback | SoME

Implementing Automatic Differentiation in Pure Python

Audience:

Tags: automatic-differentiationneural-networks

A recording of me explaining and implementing automatic differentiation in pure Python. I start with some mathematics of forward and reverse mode autodiff and then implement interleaved forward mode autodiff. Then I explain how to use Dual Numbers and implement forward mode autodiff using Dual Numbers. Finally, I construct a computation graph and perform a topological sort to achieve reverse mode autodiff. Audience: Anyone with a high-school math background can keep up. All math and code has been made accessible and explained so that an undergraduate can pick it up in one viewing.

Analytics

3 Overall score*

175 Rank

5 Votes

2 Comments

Comments

2.9

This is an amazing reference for auto-diff explained at its core. I did have to watch on 2x though because of the scale.

I think if this was broken up into a playlist it would be amazing

This video lacks any introduction to interest the viewer. It is not clear what the point of the video is or how it will be achieved.

The narration is boring and unclear. There is no editing. There is no sense of achievement or conclusion at the end.