Results 33 comments of Sergey Lisitsyn

Hey, you're right the matrix has to be self-adjoint. Isomap should produce self-adjoint matrix as well so I will have to check what's going on there.

I can reproduce some asymmetry in isomap matrix, will let you know what caused that

@coderlevi thanks for researching on that. You are right, it was somehow missed that neighborhood matrix (hence, the shortest path matrix) is not symmetrical. I think it makes sense to...

I'd keep it open until I or someone else merge some code fixing that

Hey! Sorry it took a while to respond you. The problem is that the neighborhood strategy being used in Tapkee is dead simple: it finds the nearest neighbors for each...

Hey, thanks for the report. It seems to be an actual bug. The `eval` thing resolves that indeed but it would bump the memory usage 2x for a moment. Let...

It seems that eval should be here. I just added it in ec3514033c036795ca819d3ac68b61a58d24ac53

Asked @karlnapf about that and apparently good idea is to collect feature vectors until solution stabilizes

Completely lost it! So sorry about that. Is it still relevant?

Thanks for reporting! I am not sure what caused this error yet. Have you tried other versions of eigen?