What is the optimization scheme of numpy.einsum ? Does it optimize the memory consumption or the CPU or anything else ?
I don't see this discussed anywhere. Maybe it's related to the different ‘optimize’ options, that I don't understand
Asked
Active
Viewed 171 times
0
-
Personally my experience and experience of my group mates says "Never to use np.einsum, because it is too slow. Just use element-wise operations and multiplication of matrices." – CrafterKolyan Jun 08 '19 at 21:55
-
Mostly it has to do with evaluation order when using 3 or more variables and complex combinations of indices. Look at `einsum_path` docs. – hpaulj Jun 09 '19 at 04:06
-
1Give a sample of the kinds of calculations you are using it for. – hpaulj Jun 09 '19 at 04:11
-
In general `np.einsum` is faster than element-wise operations for simple operations. Use `optimize=True` if you have GEMM's. – Daniel Jun 25 '19 at 18:33