Aaron Lun

Thanks @yixuan, using a negative value seems to do the trick. I think my input matrix is always positive semidefinite so it should be safe. It might be worth mentioning...

Circling back to this. I just tried it out on a much larger matrix, and unfortunately, it didn't work out well. After some effort, I created a reproducible example at...

Thanks both. It seems that the trick also reports the eigenvectors correctly; I'll try it out with some real data later this week.

After implementing this, and then rereading the comments above, I just realized I managed to misunderstand the math. Instead of subtracting twice the max eigenvalue from the PSD and looking...

Thanks all for the comments. Responses interspersed below. > can you elaborate a bit why this is needed? In my case, I am cross-compiling for a different architecture (specifically, Webassembly)....

Sure; now that I think about it, I actually don't need the `FetchContent` part for my projects. I just put it in as a courtesy for other users who didn't...

FWIW I was going to go for MIT for **umappp** before this discussion. Dunno if that's appropriate here, all my licensing lectures went in one ear and straight out the...

Yep, I was going to supply the NN's myself (https://github.com/LTLA/knncolle). On a tangentially related note, it would be nice to make a pure C++ port of nndescent available from that...

No worries. I also failed to make any progress as well, got distracted by https://github.com/LTLA/qdtsne.

Made a start on the initialization: https://github.com/LTLA/umappp. The most that can be said right now is that it compiles and runs.