Feature/path integral simulated annealing #72
Conversation
There was a problem hiding this comment.
Another relatively quick pass. Most comments are stylistic. I assume the C++ code has been in use for a while so I am inclined to not make structural changes. That said, can we run it through clang-format if we haven't already? And are there and C++ tests we can add?
| // add thread_local (C++11) support for MSVC 9 and 10 (py2.7 and py3.4 on win) | ||
| // note: thread_local support was added in MSVC 14 (Visual Studio 2015, MSC_VER 1900) | ||
| #if _MSC_VER < 1900 | ||
| #define thread_local __declspec(thread) | ||
| #endif |
There was a problem hiding this comment.
based on the comment, I don't think we need this anymore. Don't remove it from elsewhere (it'll make a long review even longer 😄), but we don't need to add it here.
|
I'm going to be using this code, I'll check the documentation is comprehensible! |
…ove unnecessary files. Not done: google-style docstrings and namespace all C++ code.
…ove unnecessary files. Not done: google-style docstrings and namespace all C++ code.
| @@ -1,4 +1,4 @@ | |||
| # Copyright 2018 D-Wave Systems Inc. | |||
| # Copyright 2024 D-Wave | |||
There was a problem hiding this comment.
Looks like there are no functional changes in this file. I would rather formatting changes happen as part of another PR. But certainly the copyright date doesn't need to be updated.
| @@ -1,4 +1,4 @@ | |||
| # Copyright 2018 D-Wave Systems Inc. | |||
| # Copyright 2024 D-Wave | |||
There was a problem hiding this comment.
Looks like there are no functional changes in this file. I would rather formatting changes happen as part of another PR. But certainly the copyright date doesn't need to be updated.
| @@ -0,0 +1,153 @@ | |||
| /* | |||
There was a problem hiding this comment.
Similar to the comment on the README, I don't think we want this file in the package. It will still be in the other repo though
| features: | ||
| - Add ``PathIntegralAnnealingSampler``, a SimulatedAnnealingSampler style wrapper for the dwave-pimc repository. This supports simulation of path-integral dynamics and equilibrium sampling of quantum Boltzmann distributions: https://doi.org/10.1038/s41467-021-20901-5 | ||
| - Add ``RotorModelAnnealingSampler``, a SimulatedAnnealingSampler style code for rotor model simulation. This supports a standard classical algorithm for approximation of spin dynamics: https://doi.org/10.48550/arXiv.1401.7087 |
There was a problem hiding this comment.
Ah, I was wrong about the docs failures. It is this PR, specifically it doesn't like the : in the release notes.
---
features:
- Add ``PathIntegralAnnealingSampler``, a SimulatedAnnealingSampler style wrapper for the dwave-pimc repository. This supports simulation of path-integral dynamics and equilibrium sampling of quantum Boltzmann distributions (`King et al., 2021 <https://www.nature.com/articles/s41467-021-20901-5>`_).
- Add ``RotorModelAnnealingSampler``, a SimulatedAnnealingSampler style code for rotor model simulation. This supports a standard classical algorithm for approximation of spin dynamics (`Shin et al., 2014 <https://doi.org/10.48550/arXiv.1401.7087>`_).Thanks @JoelPasvolsky for the help with this.
Add path-integral and rotor-model dimod.Samplers(). The former is a wrapper for https://github.com/dwavesystems/dwave-pimc; the latter is new code.
At the python-wrapper level code is structured to align closely with SimulatedAnnealingSampler. Under default operation the main difference is the 'classical state' on which simulated annealing is performed: binary states (
SimulatedAnnealingSampler), [discretized-]angles (RotorModelAnnealingSampler), or paths (PathIntegralAnnealingSampler, aliasSimulatedQuantumAnnealingSampler); These different states require different local update methods, and the schedule is generalized to allow variation of a transverse field in the latter cases, but otherwise the algorithms are very similar.These samplers have been carefully studied over many years as emulators of quantum dynamics and as optimization heuristics: https://doi.org/10.1038/s41467-021-20901-5