Intro to Computational Physics | Hacker News Reader

Intro to Computational Physics

mas.lvc.edu

80 pointsagumonkey27 comments

All it teaches is Euler's method (and the euler-cromer variation)? That's .. not much and pretty basic. I'd expect a lot more from a comp. physics course. Usually you see at least something like symplectic integration, solving PDEs, quadratures, and a monte-carlo simulation like a 2D Ising model. I've taken two comp. physics courses at the undergrad and grad level. definitely preferred the course where he explained the algorithm in pseudo-code on the board, and we could write it however we pleased.

I didn't even take a computational physics course as part of my physics undergrad, and I still learned about the 2D Ising model, the Monte Carlo method (which I implemented for a project), and solving ODEs with the Runge Kutta method. The Euler method is too inaccurate to be useful. The only thing it might be good for is teaching about numerical integration.

> The Euler method is too inaccurate to be useful.

You can make it as accurate as you want by shrinking the time step. :)

That's no way to go through life.

This is actually not true since rounding errors will compound.

Sure, in IEEE 754! But to call the various explicit variants of Euler useless is a bit off. Symplectic Euler is quite nice in certain applications. Hell, Newton derived a number of theorems by reasoning about explicit Euler discretizations (usually geometrically).

Yeah, it's pretty light, but look at the course number--it's a sophomore level course. Most physics students who'd be taking this course (likely as an elective or part of a focus/specialized track) are taking "Physics III" and "Calc III" concurrently, or if they're taking it in the second semester concurrently with something like undergrad classical mechanics.

They won't have been exposed to PDEs (or, at least, only tangentially through the multi-dimensional topics covered in "Calc III" or the second physics course of their freshman year) or problems reaching the level of sophistication that would make, e.g. Monte Carlo "interesting" to use (in my opinion).

Agreed; a similar course I took as an undergraduate sophomore covered all of this material in two weeks. I'll see if I can dig up some of my old notes. If the HN community has a desire for more computational physics, I'd be happy to share.

I can't speak as to why this specifically was posted, but the school is in my general area (less than an hour away) and is not particularly well regarded on the academic front. I'm not familiar with Dr. Walck so I don't mean that as a judgment on him at all, just referring to the school generally. If your focus is athletics it's probably okay despite being D3.

What if this is one guy at the school working hard to raise it's profile and you dissed the effort without evaluating it's merits?

According to Wikipedia:

------------- Consistently named to US News & World Report's "America’s Best Colleges" list under the Regional College (North) and Best Value School categories. The Princeton Review also recognized the college as "A Best in the Northeast," and Lebanon Valley College was also named on Forbes' list of "America's Top Colleges." -------------

I'm not familiar with LVC, but it seems like it's regarded as a pretty decent place to go to school.

I can tell you this is not the norm. I took and TA'd a computational physics course when I was in Uni. There was a lot more in that course. Euler's was taught but quickly went to using RK4. Lotka-Volterra (predator vs prey modeling) were discussed in depth, PDEs, monte-carlo (had students calculate pi), lots on SHO systems (common equation type in classical and quantum mechanics), various fitting functions, and even basic signal analysis. This served as the introduction to computing for our physicists too, where they could use MATLAB or Python. Above course seems lacking.

Was it a 200-level class, though?

Yep. And so is the one in this post. If there's interest I'll message the professor and post it. But it is mostly homework assignments and some bad notes. (You can do directory transversals for solutions. He doesn't care that this exists...)

Similar in spirit (but not so introductory and specific to mechanics): Structure and Interpretation of Classical Mechanics, by Sussman and Wisdom. Whole book available online here: https://mitpress.mit.edu/sites/default/files/titles/content/... .

(SICM uses Scheme -- of course! -- where this course uses Haskell. SICM is cited in the paper linked by agumonkey as an influence on the approach taken in this course.)

SICM also goes into way more physics than this course page does. Lagrangian / Hamiltonian mechanics, canonical transformations leading to symplectic integration, phase spaces, and perturbation theory.

SICM is indeed mentioned in the pdf references (I linked it below)

I find the combination of Haskell and such simple techniques quite odd. This seems more like an introduction to Haskell using some toy physics computational examples.

  - paper: https://arxiv.org/abs/1412.4880
  - source: https://twitter.com/prezcannady/status/807065697608224768

I wish courses like this where available for CS majors instead of math and physics courses. I might actually learn something.

I immediately love that this is taught in Haskell. I hadn't learned how excellent Haskell is for (non-hpc) computational physics until well into my Masters, where I began to replace my python analysis scripts.

And it's a far sight more elegant than my first Numerical Methods class, which used MatLab.

Related topic: What software do theoretical physics use when developing theories? Mathematica? Or do they just use paper and pencil?

Developing theories: pen & paper Testing theories: anything from Mathematica, Python and if things get serious: Fortan

Looks cool. Any way you can share the notes?

I'm just pushing what the tweet above said. Maybe notes/lectures will be uploaded later .. I don't know.

thanks for sharing this, looks really cool

In case you don't know, there's a book called SICM Structure and Interpretation of Classical Mechanics (by Sussman, SICP author). Similar idea, physics + computing, except it's a scheme library. Google it.

Reply on news.ycombinator.com