SciPost Submission Page
A NEAT Quantum Error Decoder
by Hugo Théveniaut, Evert van Nieuwenburg
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | Everard van Nieuwenburg |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2101.08093v2 (pdf) |
| Code repository: | https://github.com/condensedAI/neat-qec |
| Date accepted: | 2021-06-24 |
| Date submitted: | 2021-06-03 11:43 |
| Submitted by: | van Nieuwenburg, Everard |
| Submitted to: | SciPost Physics |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approach: | Computational |
Abstract
We investigate the use of the evolutionary NEAT algorithm for the optimization of a policy network that performs quantum error decoding on the toric code, with bitflip and depolarizing noise, one qubit at a time. We find that these NEAT-optimized network decoders have similar performance to previously reported machine-learning based decoders, but use roughly three to four orders of magnitude fewer parameters to do so.
Author comments upon resubmission
This new version of the manuscript has an improved presentation (notably the consistency of Fig. 2 with the ones in the appendices).
Both (pre-editorial recommendation) referees raised excellent in-depth points (for which we are very grateful), and we believe that we have addressed these points in the current version. The third post-editorial recommendation report is very motivating too. Replies to the referees will follow separately.
We have now also explicitly included a link to our github repository that contains all of the code to reproduce our results.
Thank you for your effort and patience in getting our work reviewed.
List of changes
Changes and additions to text and figures:
* Added sentences about the convergence criterion for our runs in section IV and in appendix C.
* Add link to github repository hosting all code for reproducibility
* Updated figure 2 to show the non-rotated code
* Updated text underneath Algorithm 2 to reflect that we do, in fact, expect to get a maximum likelihood decoder (Thank you @Referee #1)
* Added paragraph in Results section discussing the discrepancies between the NEAT decoder and MWPM
New citations:
* J. R. Wootton and D. Loss: https://arxiv.org/abs/1202.4316
* Meinerz et al.: https://arxiv.org/abs/2101.07285
* Chamberland and Roonagh: https://arxiv.org/abs/1802.06441
Fixed typos:
* Extra “of” in the first sentence of section IV (Thank you @Referee #1)
Published as SciPost Phys. 11, 005 (2021)
Reports on this Submission
Report
As stated already in my first report I think this is a very interesting and novel approach to decoding topological error correcting codes. The paper is very accessible, clearly structured and presented. With the modifications made in response to the referee reports I recommend that it be accepted to SciPost Physics. I agree with the assessment of the 3rd referee that this is a top 10% paper.