Implement a generic, deterministic contraction-order optimization about tensornetwork HOT 8 CLOSED

@Thenerdstation I'm interested in working on that. Could you assign me? It seems I can't assign myself

from tensornetwork.

Done. Do you wanna setup a video chat to discuss engineering requirements?

from tensornetwork.

thanks. Yes let's do that. Would someday next week be OK?

from tensornetwork.

Yep! Let's continue this over email.

from tensornetwork.

Should I be worried about trying to do this in python? Seems like a search algorithm for an NP-hard problem should at least not have to suffer from interpreter overhead! Could we use cython, for example?

from tensornetwork.

I talked to a guy at IBM about this, and they use Cpp for it. I think an initial implementation in
python is fine, but ultimately it should be ported to C/Cpp (or cython) (or small networks it might not matter too much if it's run in python)

from tensornetwork.

I agree. Let's do it in this order

Python -> Cython -> C++

We're short on headcount to keep C++ code stable right now, so I'd rather just start with python and upgrade it as usage increases

from tensornetwork.

Sounds good!

from tensornetwork.