×
Login Register an account
Top Submissions Explore Upgoat Search Random Subverse Random Post Colorize! Site Rules Donate
2

Why Runge-Kutta is SO Much Better Than Euler's Method

submitted by happytoes to mathematics 9 monthsAug 23, 2024 16:00:43 ago (+2/-0)     (www.youtube.com)

https://www.youtube.com/watch?v=dShtlMl69kY

A video about numerical methods for solving differential equations. With pretty animations. And a nice exposition of fourth order Runge-Kutta


4 comments block


[ - ] Dingo 0 points 9 monthsAug 23, 2024 17:02:46 ago (+0/-0)*

RK4-THE-WIN!

Note that you can't solve "stiff" equations using Euler's method without error or more iterations (because the slope can become near vertical in places in the solution). I think Simulink uses variable RK4 by default and it brute force solves everything within a blink of an eye.

EDIT: Watched the video, he also mentions ODE45 in MATLAB uses RK4 and RK5. Curious question (semi related): have you heard of a guy named "Carl Bender"?

[ - ] happytoes [op] 0 points 9 monthsAug 23, 2024 18:09:32 ago (+0/-0)

No. Looked and found https://en.wikipedia.org/wiki/Carl_M._Bender#Research

Sounds very advanced.

[ - ] Dingo 1 point 9 monthsAug 23, 2024 18:52:44 ago (+1/-0)

This one is about accuracy of numerical computations. He talks about "computing accurately" above "mathamatical proofs". I thought about him with your post and he talks about some neat stuff here. I thought you might like it.

https://www.youtube.com/playlist?list=PLDS6JFNmWHR3Jm93AyLRl5t-kjx3Eo33D

[ - ] happytoes [op] 0 points 9 monthsAug 23, 2024 21:23:12 ago (+0/-0)

Had a quick watch of the first three minutes of the first video.

It is very nice when you have a problem that you can solve exactly but almost no problems are exactly solvable, that is the problem and that is why it is difficult. And all the exactly solvable problems (and there are about three or four of them, that's all) um, have already been solved unfortunately and all the rest of the interesting problems that need to be solved cannot be solved exactly ...

I've also taken a peak at lecture number three where he starts off talking about other ways to make solving x^5+x=1 into a perturbation problem, so I think that lecture number one is about to turn into hard maths and its past my bed time,...