Input types for count matrix
adamgayoso opened this issue · comments
Adam Gayoso commented
We should at least accept:
- Sparse matrices
- Pandas DataFrames
- 10x mtx format
For sparse matrices, we will probably just instantly make them dense.
Jonathan Shor commented
If it seems worthwhile to fully support sparse in and out, there are some options for getting "around" PCA densifying the matrix, such as a random subsample PCA.
I'm not familiar with these approaches, so we'd have to look into them further.
Jonathan Shor commented
Likely use sklearn.utils.check_array
.