Wow, love this, thanks, such a broad subject that interdisciplinary opportunities must abound. Work like this combined with the huge amount of effort in open source work to move quickly from examples in papers to running code makes for an exciting future.
Visualising this volume of data on a mobile device is always going to be challenging, personally I'm hoping that moves in AR, Rollup/Folding Displays and projected screens will shift things towards larger screens. Mobile phones are cool, but you lose a lot of interaction potential moving from a 27" screen to 6".
Hmm. Says "shift-click" to open poster on NeurIPS site -- you'll have to be logged in to the NeurIPS virtual conference, and the authors will have had to remember to upload their poster PDF.
If you want the paper pdf, you can "alt-click" to open paper on OpenReview.
Yeah it doesn't work that well on mobile. It takes a LOT of blood, sweat, and tears to get these D3 visualizations to be mobile friendly and this is just a fun project after all.
It actually works pretty well for me in iOS Safari — the main problem is that the visualization is sandwiched between the two sidebars, so it's about half an inch wide. If the three divs stacked vertically on small screens, that would be a huge improvement!
We regularly use pygraphistry to generate /import => viz 100k+ entity embeddings on mobile fine: https://github.com/graphistry/pygraphistry , should be pretty easy to redo / scale . WebGL is impressively capable on even old smartphones nowadays !
More fun, in umap mode, by default, it also shows the top-n similarity edges between each entity, so you get interactive graphs that you can interactively filter + recluster + restyle on diff data attributes, vs just the typical x/y scatter plot
(Am on mobile, so haven't been able to load the original...)