Leland McInnes
Leland McInnes
Yes, that's the one. I got basic code hammered out, but hadn't fixed all the bugs. I also chatted with Matt Rocklin and it looks like the proposed solution would...
I didn't get it working to a point of pushing it up anywhere. I have a colleague who may be picking up this work, so I might try and loop...
Currently this is still, at best, an "in the works" issue. The first real hurdle to be overcome, the nearest neighbor search, is looking good. There is a paralllel NN...
At present it looks like this is something that might be possible in the v0.5 to v0.6 time-frame. That's not soon, but it may yet happen.
Depending on what exactly you are looking for, this may now be available (in preliminary form) in the 0.5dev branch. To make it work you would use the new, special,...
That's definitely a bug -- as you say, it should treat that as no relation. I presume you have a dataset you want aligned except at some intermediate slice where...
I think a lot of this is beyond what the basic UMAP approach can do as it is essentially transductive. For those who really need this I would suggest your...
Part of the catch here is that you are in the "small data" regime of UMAP, and the transform is (relatively) less efficient in this case. The transform method is...
If you have a lot of data then fit is always the best option. There is a shortcut if you are transforming precisely the training data (hence the speed in...
Yes, that is indeed the case -- we are storing an index for approximate nearest neighbor search, and that involves storing the original data.