BiVector: Community for Geometric Algebra for CGI, Vision and Engineering | Hacker News Reader

BiVector: Community for Geometric Algebra for CGI, Vision and Engineering

bivector.net

65 pointsadamnemecek13 comments

The community is brand new. It's managed by Steven De Keninck (enkimute), the author of the ganga.js framework (https://github.com/enkimute/ganja.js).

Check out some examples

https://enkimute.github.io/ganja.js/examples/coffeeshop.html...

Why isn’t geometric algebra more popular when it is described glowingly as “the most powerful language for mathematical physics”? (Honest question)

No idea, I freaking love it and no one would confuse me for a mathematician or physicist. (I'm using Versor in C++.)

Because there's a lot of preexisting mathematical physics work (ie legacy code) that uses basic vector algebra, so you have to know that, whereas you can get away with not grokking geometric algebra, so plenty of people do.

Apart from the exterior algebra part, there is little added value to using geometric algebra compared to other representations.

My guess is that most applications don't really need exterior algebra, or at least they would not benefit much from the nice unified representation it provides.

Hi, I'm the developer of Grassmann.jl, Actually, it has a lot more value to provide than what is taught in school. It never caught on because it was historically difficult to understand without computer algebra to help build intuition.. so the only people who understood it are the ones bold to explore mathematics deeply enough to arrive at this system and utilize it.

> there is little added value to using geometric algebra compared to other representations

This is not right in my experience. When trying to solve geometry problems I find that working out the details in terms of the GA formalism typically saves big piles of work compared to other formalisms (such as matrices, differential forms, trigonometry, analytic geometry, ...).

The biggest features are (a) access to a concept of “multivectors”, and (b) access to a concept of vector division.

Mathematicians use these concepts all the time in the planar setting by pretending that points and vectors are just funny kinds of complex numbers, and eliding the semantic differences between these different types of objects. Similarly for the use of quaternions in 3D (though this is less common nowadays than complex numbers).

But multivectors and vector division are very powerful tools in more general settings.

What sometimes ends up happening is I try to solve a problem using my (more extensively trained) understanding from standard undergraduate math courses and other standard textbooks, flail around with a bunch of horrible fiddly algebra for a while, then decide to redo my work in terms of GA and end up with like 4–5 lines of simple manipulation replacing a page of messy work.

And I am by no means an expert. There are many extremely convenient (multi)vector identities which I have not properly learned and end up laboriously working out for myself while solving geometric problems. If I had spent more time doing guided exercises I am sure I could be still much more efficient.

I completely agree: for pen/paper computations it is probably hard to beat.

I was more referring to computer applications, where once you've coded data structures for point/lines/planes/etc you don't really care how they're implemented.

I am also a fan of the way it simplifies things for me. There are efforts to introduce it earlier in education as an alternative to the normal way these things are taught. One uses Geogebra to build up from simple to more complex concepts:

https://www.geogebra.org/m/qzDtMW2q

Also see slides[0] from a recent course at SIGGRAPH2019 that Steven De Keninck and Charles Gunn held.

[0]: https://slides.com/enkimute/siggraph/

Do you recognize that BiVector are Steven De Keninck and Charles Gunn ?

Hi (Steven here) - biVector.net is an effort from active researchers from Cambridge, UPEM Paris, UVA Amsterdam, and a couple of others. (as well as the developers for some modern, actively maintained GA libraries).

Nice visualizations on the homepage! Cool that one is interactive, but some hijack the scroll event and the right-click menu.

Also the super light contrast on the de-emphasized text makes a lot of the content hard to read.

Reply on news.ycombinator.com