Automatic Differentiation: Differentiate (almost) any function

Automatic Differentiation is the backbone of every Deep Learning Library.
GitHub: github.com/tgautam03/jac
Music: No One Is Perfect by HoliznaCC0 (freemusicarchive.org/music/ho...)
00:00 - Recap
00:30 - Topics Overview
01:00 - Finite Differences
02:40 - Automatic Differentiation (Forward Pass)
04:28 - Local Gradients
05:30 - Backward Pass
07:38 - Conclusions
#3blue1brown #ml #machinelearning #neuralnetwork #ai

[@CodeX-hv1sf]

wow this is soo neat i loved the content

[@sarathkumarmulugurthi9206]

Great video and great effort. Thank you very much. Now i am clear on the concept. But i think there might be a small mistake. dL3/dL2 = n right? because you are differentiating a sum of L3 (n times?) rather than L3 alone itself. I might be wrong.

[@harshsharma03]

This is the animation tool that 3b1b uses right?

[@0Mean1Sigma]

Yes. It’s available as an open source library called Manim.

[@harshsharma03]

@@0Mean1Sigma Thanks. Your videos are great btw, definitely keep them coming!

[@0Mean1Sigma]

Thanks a lot. I plan on posting frequently and have topics other than ML on my mind too.