Yuchen Li

> Different embedding seed perhaps? If you have the same uwot, Seurat or umap-learn then that is the only thing I can think of. Hi, Kristof. I don't think it...

Hi, @timoast Sorry for the long-time delay! It seems that the different UMAP embeddings are derived from the different PCA outputs, which due to `irlba::irlba()`. I replaced it with `RSpectra::svds()`...

There must be some mistakes during converting `.md` to pdf and all the outputs of cells were lost! I regenerated the `.png` from the original notebook. 

Hi, @timoast , I found another potential source of non-reproduciblility, that the variable features output from `FindVariableFeatures()` may be in different orders on different machines, even the feature sets were...