NumPy Matrix Deprecation about geometric HOT 6 CLOSED

Thanks. I only used matrices when matrix multiplications were actually needed, and converted them back shortly afterward. Otherwise it became too confusing to remember whether a variable somewhere was an array or a matrix. It should be straightforward to make the changes.

from geometric.

Thanks for the notification. This seems troublesome ... is there an easy way to patch the code while making sure all the linear algebra is still right?

from geometric.

Unfortunately not really. Changes like this are normally validated via a full test suite. You can also attempt to compare operational pathways directly by leaving the matrix operations in place, adding an array routine, and using a assert np.allclose for each call until you are satisfied the code is producing the same results.

from geometric.

You may be interested in the Python 3.5 @ operator. Otherwise np.dot or np.linalg.multi_dot are good alternatives.

from geometric.

Thanks for letting me know about this - it seems very nice, though I might prefer to use numpy.dot for now for Python 2.7 compatibility reasons.

from geometric.

Marking this as resolved, thanks again for the heads-up.

from geometric.