Dimension mismatch

[BJWiley233]

Describe the bug A clear and concise description of what the bug is.

Here's a quick checklist in what to include:

• [ ] Include a detailed description of the bug or suggestion
• [ ] pip list or conda list of the environment you are using (please attach a txt file to the issue).
• [ ] deeptime and operating system versions
• [ ] Minimal example if possible, a Python script, zipped input data (if not too large)

I am not sure what type of multiplication we want to do here , i.e. is it element-wise multplication, or are we doing matrix multiplication, but I get an error when it tries to multiply pi0[:, np.newaxis] which for me is a numpy array with shape (11343, 1) and sums to 1 with P0 which is a csr_matrix of shape (11343, 11343) and each row sums to 1.

pi0[:, np.newaxis] * P0

gives ValueError: dimension mismatch

So instead which do we want?

V0 = P0.multiply(pi0[:, np.newaxis])  # Element-wise multiplication

which will be shape (11343, 11343) then each row will then add up to each corresponding entry in pi0[:, np.newaxis]

or matrix-multiplication which gives the sum of each row as a 1D vector, i.e. not a row but the row sum which again totals to 1.

pi0[:, np.newaxis].T * P0
# or
pi0[:, np.newaxis].T @ P0

(pi0[:, np.newaxis].T * P0).shape is (1, 11343)

[BJWiley233]

Maybe also is P0 is never supposed to be sparse?

[clonker]

it should be possible for P0 to be sparse in this case, but I don't think it was ever properly tested. In the dense case this is element-wise (and broad-casted) multiplication, not matrix multiplication. This should be fixed in any case. For the time being, if your memory allows it, you could try using dense matrices here.