プラットフォーム検査
このビューでは、プラットフォームの公開機能の豊かさ、一般的な最適化フレームワーク種別に 対する公開安全な能力比較、分布距離較正ラベル、そしてそれらの主張の背後にある成果物の 来歴を検査できます。下記のすべては公開安全な派生エビデンスでビルド時にインライン化されて います——私的エビデンスなし、ライブ計算なし、性能主張なし。
These tables compare platform capabilities, features, and artifacts. They are not direct performance claims; performance evidence lives in the statistical analysis exports.
All rows are public-safe derived data (derived-public-safe); each carries an explicit evidence-source label. No private evidence, live compute, or unpublished claim is surfaced.
Feature richness
| Feature | Present | Evidence source | Note |
|---|---|---|---|
| benchmark catalog with characterization | yes | artifact-inspected | — |
| named-solver registry | yes | artifact-inspected | — |
| multi-objective and constraint accounting | yes | artifact-inspected | — |
| distribution-distance calibration labels | yes | documented | — |
| non-parametric statistical analysis | yes | artifact-inspected | — |
| reproducible download bundles | yes | artifact-inspected | — |
Comparator-framework comparison
| Capability | Comparator class | DispatchAtlas support | Comparator support | Evidence source |
|---|---|---|---|---|
| exact optimal scheduling | general CP/MILP toolkit | optional exact adapters plus native bounded DP | first-class exact solvers | documented |
| population metaheuristics | general metaheuristic library | registry of constructive and metaheuristic solvers | metaheuristic operators | documented |
| many-objective optimization | general many-objective library | NSGA-III with structured reference points | many-objective algorithms | documented |
Distribution-distance labels
| Bridge profile | Status | Divergence | Threshold | Calibration | Evidence source |
|---|---|---|---|---|---|
| cloud-edge-capacity | synthetic | — | 0.25 | not calibrated | documented |
| workflow-dag | synthetic | — | 0.25 | not calibrated | documented |
| machine-scheduling-unrelated | synthetic | — | 0.25 | not calibrated | documented |
| rcpsp-renewable | synthetic | — | 0.25 | not calibrated | documented |
Artifact provenance
| Artifact | Kind | Present | Reference | Evidence source |
|---|---|---|---|---|
| benchmark-catalog.json | portal-data | yes | assets/portal/benchmark-catalog.json | artifact-inspected |
| solver-capabilities.json | portal-data | yes | assets/portal/solver-capabilities.json | artifact-inspected |
| results-preview.json | portal-data | yes | assets/portal/results-preview.json | artifact-inspected |
Limitations
- Distribution-distance bridges ship as synthetic profiles: no trace-backed reference set is bundled, so divergence stays unscored and each row reports its calibration status only.
- Bundled result previews and interactive chart series carry smoke-development evidence; no reference-grade campaign evidence ships in the portal downloads.
ナビゲーション
- ベンチマークカタログ は機能行と分布距離行を 裏付けます。
- ソルバーレコメンダー と 結果ポータル はここで要約された能力エビデンスを保持します。
- 比較 は導入志向のビューを枠付けます。