How do I effectively normalize data in large sparse matrix?

[WiedenWei]

Hi, all

I have large sparse matrix (single cell rna count matrix, typically 40000*10000) in hand, and I am new to nalgrebra, the doc is hard to read for me, how do I normalize data in large sparse matrix effectively. For example, counts for each column are divided by the total counts for that column and multiplied by a scale factor, then natural-log transformed. Are there any buildin routines for that, instead of iteration. Or anyone has a clue? Thanks in advance!

[Andlon]

Hi @WiedenWei,

assuming you're using a CscMatrix, then you could do something like this:

for col in matrix.col_iter_mut() {
    // Number of non-zero entries in column
    let col_nnz = col.nnz();
    for entry in col.values_mut() {
        *entry =  // divide by col_nnz, multiply, log-transform and so on
    }
}