Question

There does not seem to be a method in scipy.sparse which gives the minimum of a sparse matrix. In particular, I seek the minimum of the columns.

No method appears in the docs and numpy minimum does not apply. If X is a sparse matrix, X.min() also throws the error: *** AttributeError: 'module' object has no attribute 'min'.

Surely this must be something people use. How is this done?

Answer 1

With CSR/CSC matrices, use

def min_sparse(X):
    if len(X.data) == 0:
        return 0
    m = X.data.min()
    return m if X.getnnz() == X.size else min(m, 0)

To do this per row or column, you can map this over X.getrow(i) for i in X.shape[0] or X.shape[1].

But you’re right, this should be a method.