Cell Type Classification Datasets

[nc1m]

Hi,

I'm trying to reproduce your cell type classification results on the Multilpe Sclerosis, Myeloid and hPancreas datasets.

In what form are these datasets, since they contain float values I guess they are not raw counts? Have these 3 datasets already been normalized and logarithmized, thus only need to be binned?

[nc1m]

@subercui I'm asking because if you preprocessed the datasets with scanpys normalize_total and log1p, applying numpys expm1 (reverse of log1p) and summing the values for each cell should result in a normalization constant, which is not the case. Thus I'm wondering if I'm missing something.

In #174 it is mentioned that the ms dataset is also normalized in the Tutorial_Annotation.ipynb