保存先:
| 主要な著者: | , |
|---|---|
| フォーマット: | Recurso digital |
| 言語: | 英語 |
| 出版事項: |
Zenodo
2026
|
| 主題: | |
| オンライン･アクセス: | https://doi.org/10.5281/zenodo.19689983 |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
目次:
- <p># Changelog</p> <p>All notable changes to the SRP-HUBO artifact are documented in this file.<br>The format is based on Keep a Changelog; the artifact uses<br>[semantic versioning](https://semver.org).</p> <p>## [2.0.0] - 2026-04-22 (TQE major-revision response)</p> <p>### Added<br>- `src/srp_hubo/` library subpackage (Phase-4 refactor):<br> - `Solver` ABC + `register_solver` decorator + 14 built-in adapters<br> under paper-stable names (`bals`, `sa`, `tabu`, `ga`, `pt`, `exact`,<br> `continuous`, `milp_round`, `bsb`, `qaoa_vanilla`, `qaoa_warm`,<br> `qaoa_bals_warm`, plus two BALS ablations).<br> - `check_solver()` conformance harness with 9 checks<br> (instantiability, return type, shape/dtype/range, `f_lambda`<br> consistency, runtime sanity, determinism, scale-invariance of the<br> gap).<br> - Draft-07 JSON schemas for serialised instances and result<br> certificates.<br> - `srp-hubo-lint` CLI (schema + semantic validation) and<br> `srp-hubo-check` CLI (conformance driver).<br> - 25-test pytest suite covering objective math, gap scale-invariance,<br> BALS, the conformance harness, and schema round-tripping (< 2 s).<br>- HiGHS / CBC open-source MILP backend in `src/miqp.py`; backend is<br> auto-resolved at call time and reported through<br> `SolverResult.meta["backend_used"]`. Gurobi remains opt-in.<br>- `docs/developer.md`: architecture, solver-implementation guide, CLI<br> reference, deferred-items list.<br>- Dolan-Moré performance profile (`scripts/27_dolan_more.py`,<br> `results/dolan_more_*.csv`, §VI-G of the paper).</p> <p>### Changed<br>- Penalty / Holm-Bonferroni family expanded from 10 pairs to 11 pairs<br> (added BALS-warm QAOA vs. BALS).<br>- Proposition 2 "Conditional scale invariance" replaces the earlier<br> unconditional statement; ε-floor regime boundary is now explicit.<br>- Paper-visible artifact tag bumped from `srp-hubo:v1.0` to<br> `srp-hubo:v2.0`; new Zenodo DOI minted for v2.0 (see `CITATION.cff`).</p> <p>### Fixed<br>- `jsonschema` draft-version compatibility: schemas now declare<br> Draft-07 so the bundled validator works under `jsonschema>=3.2`.</p> <p>## [1.0.0] - 2026-04-19 (TQE submission)</p> <p>### Added<br>- Four synthetic benchmark regimes (gaussian / heavy / skewed / stress)<br> generated by a single-factor model with configurable skew and excess<br> kurtosis, in `src/instances.py::build_synthetic_panel`.<br>- `scripts/19_bootstrap_variance.py`: 10k-resample bootstrap CIs on<br> median gap + BALS-warm non-monotonic-p diagnosis.<br>- `scripts/20_q_grid.py`: full $5\times5$ $(q_2, q_3)$ stress grid.<br>- `scripts/21_mps_scaling_deep.py`: MPS scaling to $p \le 3$,<br> $\chi \le 128$, $n \le 24$.<br>- `scripts/22_budget_parity.py`: compute-budget Pareto sweep for<br> BALS/SA/GA/bSB/QAOA.<br>- `scripts/23_run_synthetic_benchmark.py`: runs the full solver ladder<br> on the four synthetic regimes.<br>- `scripts/24_tqe_statistics.py`: Wilcoxon signed-rank tests with<br> Holm-Bonferroni adjustment; rank-biserial effect sizes; Wilson-score<br> exact-hit CIs.<br>- `scripts/25_tqe_artifacts.py`: TQE-specific figures and LaTeX macros.<br>- Per-solver `n_evals` instrumentation in BALS/SA/GA/bSB so matched-<br> evaluation-count comparisons are possible.<br>- Dockerfile, pinned `requirements.txt`, CITATION.cff, LICENSE, this<br> CHANGELOG.</p> <p>### Changed<br>- `paper/main_tqe.tex`: new journal-length paper targeting<br> IEEE Transactions on Quantum Engineering.</p> <p>## [0.9.0] - 2026-04-15 (QCE26 RESP revision)</p> <p>Initial revision for the IEEE Quantum Week conference track. See<br>`paper/main_resp.tex` for the submitted version.</p>