Dmitry Kobak

> I think it'd be interesting to have it as an option (while the default remains as is) Hi @Muennighoff, I'm not quite sure what you mean here. To have...

@danielhanchen Thanks for a quick reply! 1. Yes! Strongly agree with using PCA init by default even though `sklearn` does not. The documentation here https://docs.rapids.ai/api/cuml/stable/api.html?highlight=tsne#cuml.TSNE only mentions random init as...

Yes, exactly!

Hi, Giovanni, thanks for reaching out! We are planning to submit a PR to scanpy, perhaps even next week. Are you one of the scanpy developers? I have some questions:...

Thanks @giovp. This is very helpful. What is the meaning of "total" in `normalize_total`? Does it mean normalize by the "total" read/UMI counts per cell? If so, then I think...

Thanks for joining in @LuckyMD. So if I understood correctly, you agree with having two new functions ``` sc.pp.pearson_residuals(adata) sc.pp.highly_variable_genes(flavor="pearson_residuals") ``` right? > Would it be possible to not use...

Replying to both comments above: > normally we would put functions like this in scanpy.external That's up to you guys. We can make `normalize_pearson_res` and `highly_variable_genes_pearson_res` and put both into...

Very good point about rank-1 \mu matrix representable this way in `LinearOperator`. That is pretty cool. However, my intuition is that it won't work with the denominator, because it's not...

@LuckyMD Hmm. We could simply call this function `normalize_pca_pearson_res()` rather than `pca_pearson_res()`, because indeed normalization happens inside. So this might reduce confusion. Then the application pipeline would be ``` sc.pp.highly_variable_genes(adata,...

@jlause Clarification: > The downstream function then does the subsetting internally. Size of adata.X changes. What I meant is that this would ONLY happen when the user calls `normalize_pearson_res` with...