Understanding scipy.minimize part 1: The BFGS algorithm
Vložit
- čas přidán 19. 05. 2023
- A description of how quasi Newton algorithms in general, and in special the BFGS algorithm work. Animations are made with the manimce library.
Sources:
* Nocedal & Wright: Numerical Optimization Ch. 6 (which also presents a SR1 trust region method)
* Dennis & More: Quasi-Newton Methods, Motivation, and Theory, SIAM Review, Vol. 19, No. 1, 1977 (describing the PSB method. The video is based mostly on the derivation in section 7 of this paper.)
The actual update formulas of BFGS are not included in the video. These can be found in both sources as well as e.g. Wikipedia. - Věda a technologie
Ridiculously underrated channel, high quality visuals and great explanation.
Awesome video! Incredible content quality for such a small channel!
Nice video! There are plenty of videos on Newton-Raphson methods but I could not find any visual explanations of quasi-Newton methods until I found yours.
Thank you!
awesome visualization. thanks
Awesome content! Keep up the good work!
Great video! Nice visuals and easy to understand
Thank you for making this video!
Ubelivably excellent.
Amazing content
thank you, it was very helpful !
Thanks a lot for making this video. :-)
beautiful
Really helpful!
Really nice video, thank you! Just a really minor correction to 04:15... due to the symmetry you need to compute slightly more than half of the values, not exactly half of the values.
Thanks! You are of course completely right with your correction!
Thanks for the video! One comment: even though you discuss the importance of convexity later on, your claim "in higher dimensions, finding the minimum of a quadratic function is very easy" is misleading ;)
"Sci-py" as in "Sci" in Science ;)
Thanks! I knew the meaning of the "Sci" but somehow never connected this to the prononciation :-/ (And the prononciation is literally on the first page of the documentation :D)
@@FolkerHoffmann lol, all good. Thanks for the great content!