Intuition behind reverse mode algorithmic differentiation (AD)

  Рет қаралды 16,540

Joris Gillis

Joris Gillis

Күн бұрын

By far not a complete story on AD, but provides a mental image to help digest further material on AD.
For a bit more context, how AD can be used to compose full Jacobians, see • High-level view of AD ...
Created with the help of LIMEL (www.kuleuven.b..., sponsered by MECO ( www.mech.kuleu... )

Пікірлер: 22
@wiczus6102
@wiczus6102 3 жыл бұрын
That smooth fade in at start
@plouf1969
@plouf1969 5 жыл бұрын
Setting aside the interesting topic, I was amazed by your writing things from 'behind' some class wall and it not appearing mirror-imaged :)
@MarcelloZucchi91
@MarcelloZucchi91 5 жыл бұрын
I think he writes normally then the video is mirrored before uploading. In fact he appears to be left handed here, whereas he seems right handed in other videos.
@navsquid32
@navsquid32 5 жыл бұрын
It would be impossible to reverse the images of the video so that the writing appeared normally.
@EvanZamir
@EvanZamir 5 жыл бұрын
@@navsquid32 What if he is on one side of a piece of glass writing and I am on the other side of the glass taking the video of him? I can then do a mirror operation on the video.
@navsquid32
@navsquid32 4 жыл бұрын
@@EvanZamir It was sarcasm, boomer.
@EvanZamir
@EvanZamir 4 жыл бұрын
Matt Williams Clearly didn’t come across as sarcasm to me, but assuming it was, did you feel like I was insulting you? Why would you respond with derogatory name calling when I was earnestly trying to provide some useful commentary? This is what is wrong with social media. You need to do some introspection.
@Horivision
@Horivision 3 жыл бұрын
You should have given an example with variables defined; this made it more complex. What's the p1 and p2 arguments all about
@queenieluo4189
@queenieluo4189 Жыл бұрын
Yes, I also wonder what are p1 and p2?
@gzitterspiller
@gzitterspiller 4 жыл бұрын
I always found it easier to work always on the tree and then just work it with graphs algorithms.
@vijaypalmanit
@vijaypalmanit 3 жыл бұрын
you have unnecessarily made it difficult, it is easy although....
@yifanwang7696
@yifanwang7696 5 жыл бұрын
at 6:17, why is (partial f4)/(partial P1) not partial (f4/partial P2)?
@yifanwang7696
@yifanwang7696 5 жыл бұрын
oooh, I saw you corrected it later at 7:18, sorry
@andyborjigin9699
@andyborjigin9699 9 ай бұрын
What are the p1 & p2 here? Are they f1 & f2? Or something else?
@gabrielmccartney7975
@gabrielmccartney7975 2 жыл бұрын
Can we use dual numbers for integration?
@larrylaw3693
@larrylaw3693 4 жыл бұрын
Great explanation, thanks!
@zephyrpenoyre2714
@zephyrpenoyre2714 5 жыл бұрын
just to check my undertsanding - are you writing += (i.e. take the previous value and increment) or plus-minus in the second half of the video (\pm in LaTeX)? I presume it's the first, but your notation is a little hard to track. Maybe a plus with a double underline would be clearer? Otherwise lovely video! Look forward to seeing more
@Banefane
@Banefane 3 жыл бұрын
Thank you very much!
@Mathomla
@Mathomla 4 жыл бұрын
Thanks for the clear explanation!
Automatic Differentiation.
12:42
Matt Yedlin
Рет қаралды 13 М.
IPOPT on steroids: fatrop solver in CasADi
19:40
Joris Gillis
Рет қаралды 2,2 М.
Миллионер | 1 - серия
34:31
Million Show
Рет қаралды 1,7 МЛН
Incredible: Teacher builds airplane to teach kids behavior! #shorts
00:32
Fabiosa Stories
Рет қаралды 9 МЛН
The Joker wanted to stand at the front, but unexpectedly was beaten up by Officer Rabbit
00:12
Самое неинтересное видео
00:32
Miracle
Рет қаралды 2,9 МЛН
What is Automatic Differentiation?
14:25
Ari Seff
Рет қаралды 111 М.
What is Jacobian? | The right way of thinking derivatives and integrals
27:14
Automatic Differentiation
19:33
Computer Graphics at TU Wien
Рет қаралды 10 М.
Algorithmic Differentiation 1
40:32
BYU FLOW Lab
Рет қаралды 2,1 М.
Tutorial on Automatic Differentiation
6:01
Matt Yedlin
Рет қаралды 10 М.
Vector and matrix derivatives
12:39
Herman Kamper
Рет қаралды 33 М.
Reverse Mode Automatic Differentiation
26:12
Nathan Sprague
Рет қаралды 3,5 М.
Simple reverse-mode Autodiff in Python
15:52
Machine Learning & Simulation
Рет қаралды 2,4 М.
The Boundary of Computation
12:59
Mutual Information
Рет қаралды 1 МЛН
Automatic Differentiation: Differentiate (almost) any function
8:41
Миллионер | 1 - серия
34:31
Million Show
Рет қаралды 1,7 МЛН