n_components parameter for t-SNE

scverse / scanpy

Single-cell analysis in Python. Scales to >1M cells.

https://scanpy.readthedocs.io

BSD 3-Clause "New" or "Revised" License

1.9k stars 597 forks source link

n_components parameter for t-SNE #1435

Open plankter opened 4 years ago

plankter commented 4 years ago

[x] Additional function parameters / changed functionality / changed defaults?

Could you please add n_components parameter for tSNE embedding? sklearn t-SNE implementation allows to set this parameter. At the moment even setting n_pcs to 3 and use_rep to three-dimensional X_pca produces two-dimensional output.

jhrdt commented 4 years ago

Hej,

I would like to implement this (n_components parameter).

If I am not mistaken, https://github.com/theislab/scanpy/blob/master/scanpy/tools/_tsne.py#L12 is the place where such an extra argument should be added.

MulticoreTSNE also supports this parameter (See: https://github.com/DmitryUlyanov/Multicore-TSNE/blob/f30b34041a24bce7d2b02bb93475b0462626d5c8/MulticoreTSNE/__init__.py#L51).

jhrdt commented 4 years ago

I am not interested anymore.

stela2502 commented 3 years ago

Is that really STILL not working?!!?

flying-sheep commented 3 years ago

Hi! The reason this is still not working is because everyone here is a volunteer.

The change is probably relatively small: How about you give implementing it a shot if it’s important to you?

stela2502 commented 3 years ago

If I would have time for that probably. But I am no Python programmer and am only a user of the tool. On the other hand scanpy is quite hopeless with 1e+6 cells. That is why I even tried whether TSNE would work. But if that does not even support 3D I might just look for more capable SingleCell analysis tool. Probably give Scarf a shot.