Performance issue in python with nested loop

Question

I was able to improve a code written in python a lot with numpy because of the dot product. Now I still have one part of the code which is still very slow. I still don\'t unders

Answer 1

Before worrying about multiprocessing, try using what plain numpy offers.

First, make sure ws are numpy arrays, not lists of some such. Then

cv1 = numpy.ndarray.sum(numpy.absolute(numpy.subtract(ws[x1], ws[x2])))
if cv1 == 0:
   f11 += 1

becomes f11 = np.nonzero(ws[x1] == ws[x2]).

Do the same for the rest of the code, and you'll be able to see more structure: np.product is just * and so on.

Then, re[x1][x2][x3] is not how you normally index numpy arrays, use re[x1, x2, x3]. This alone would save you quite a bit of time and memory allocations.

Once this is done, look if you can actually vectorize the expressions, i.e. use numpy all-array operations instead of plain python loops (it's likely that you can, but it's too hard to see with the current state of the code snippet).