data science
* yeah i guess pytorch or dask works too. RIP Torch7 you will be missed.
🇳🇱
data science
@icedquinn Force directed spring systems are notoriously inefficient. Maybe you should try my the hyperassociative map I invented. It works very similarly, as its a psudo-force directed algorithm, but is significantly more efficient.
re: data science
i ended up with springs because there is a very specific requirement that the graph nodes end up closer to some and farther from others http://iceworks.cc/z/ckjxo2z5r00001dis1dsfftl2.html
i do a little abusing of hook's law.
re: data science
@icedquinn yes my algorithm will have the same property as springs except be far more efficient at getting you there.
re: data science
@icedquinn The link I gave you has a video of it in action. The links arent visible in the video but you can see how it immediately aligns without the oscilationa nd dampening problem of springs, but with general the same final position as you'd expect from a spring algorithm.
re: data science
@icedquinn yup, thats the same sort of use case I had when i invented HAM. While it can be used for the visual side of things, of course, its real purpose was to cluster similarly connected nodes together.
re: data science
the graph itself is just a transitory mechanism to position the hot bits near or far from one another semantically.