Skip to content

Finalize GSQCA release readiness fixes #4

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
t-yamsaki merged 1 commit into from
Jul 1, 2026
Merged

Finalize GSQCA release readiness fixes #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 10 additions & 9 deletions PROVENANCE_AUDIT.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,8 @@ Scope:
- `src/qca/sweep/threshold.py`
- `src/qca/minimizers/`
- `src/qca/mlqca/`
- the former private MixedQCA project used as the starting point for PyQCA
- the former private first-party predecessor project used as the starting
point for PyQCA

This is a source-provenance review for release preparation. It is not a legal
opinion.
Expand All @@ -22,8 +23,8 @@ notice in `THIRD_PARTY_NOTICES.md`.
the PyQCA minimizer implementation.
- The mlQCA implementation uses the published protocol and public package API
as methodological references without translating R source code.
- The former private MixedQCA project was authored by the PyQCA copyright
holder and introduces no third-party licensing requirement.
- The former private first-party predecessor project was authored by the PyQCA
copyright holder and introduces no third-party licensing requirement.

## Materials Reviewed

Expand Down Expand Up @@ -83,7 +84,7 @@ Both implementations:
PyQCA differs materially in implementation:

- Python/pandas dataclasses replace R data frames and S3 result classes;
- PyQCA invokes its own `MixedQCA` and minimization backends rather than
- PyQCA invokes its own `GSQCA` and minimization backends rather than
`QCA::truthTable()` and `QCA::minimize()`;
- validation, exception handling, result storage, plotting, and Markdown
generation are independently implemented;
Expand Down Expand Up @@ -120,11 +121,11 @@ line-by-line control flow were identified in the PyQCA minimizer modules.
Quine-McCluskey and set cover are published algorithms and were implemented in
PyQCA using Python-native data structures.

## MixedQCA Origin
## First-Party Predecessor Origin

The former MixedQCA project was private and authored by the same copyright
holder as PyQCA. It is first-party code and therefore does not require a
third-party notice.
The predecessor project was private and authored by the same copyright holder
as PyQCA. It is first-party code and therefore does not require a third-party
notice.

## mlQCA Findings

Expand All @@ -133,7 +134,7 @@ modules define PyQCA-specific configuration, validation, result contracts, an
optional XGBoost backend, feature-importance summaries, split-threshold
aggregation, crisp calibration, csQCA evaluation, condition-combination
search, Pareto analysis, fuzzy-anchor proposals, fsQCA calibration, and anchor
sensitivity adapters. MixedQCA integration, Markdown reporting,
sensitivity adapters. GSQCA integration, Markdown reporting,
visualization, and reproducibility metadata are also independently
implemented with PyQCA-native components. P6 adds stratified cross-validation,
stratified bootstrap with out-of-bag evaluation, and aggregate feature and
Expand Down
36 changes: 20 additions & 16 deletions README.ja.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ csQCA、mvQCA、fsQCA、混合型QCAに加え、XGBoostによる条件選択と

## ドキュメント

Sphinxドキュメントには、インストール、Core/Mixed QCA、calibration、
Sphinxドキュメントには、インストール、Core/gsQCA、calibration、
minimization、感度分析、レポート、mlQCA、API referenceを収録しています。
Read the Docsで公開できる構成です。

Expand All @@ -35,7 +35,7 @@ Read the Docs projectを接続するまでは、ビルド後に
## 特徴

* **Core QCA**: `CSQCA`, `MVQCA`, `FSQCA`
* **MixedQCA**: crisp / multi-value / fuzzy 条件を統一スキーマで扱う高レベルインターフェース
* **GSQCA**: crisp / multi-value / fuzzy 条件を統一スキーマで扱う高レベルインターフェース
* **Minimization backends**: `standard`, `qmc`, `set_cover`, `greedy_set_cover`, `exact_set_cover`
* **Calibration**: crisp calibration, piecewise fuzzy calibration, logistic fuzzy calibration
* **ThresholdSweep**: ThSQCA に準拠した outcome / condition / grid / dual threshold sweep
Expand Down Expand Up @@ -145,14 +145,14 @@ model = FSQCA(
result = model.fit(consistency_cutoff=0.8, coverage_cutoff=0.1)
```

### MixedQCA
### GSQCA

```python
import pandas as pd

from qca import MixedQCA
from qca import GSQCA

mixed_df = pd.DataFrame(
gsqca_df = pd.DataFrame(
{
"case": [f"c{i}" for i in range(1, 9)],
"policy": [1, 1, 1, 0, 1, 0, 0, 0],
Expand All @@ -162,8 +162,8 @@ mixed_df = pd.DataFrame(
}
)

model = MixedQCA(
data=mixed_df,
model = GSQCA(
data=gsqca_df,
case_id="case",
outcome="Y",
conditions=["policy", "region", "capacity"],
Expand All @@ -177,6 +177,10 @@ model = MixedQCA(
result = model.fit(minimizer="set_cover")
```

`GSQCA` は PyQCA が文書化する操作的定義に基づく generalized-set
workflow です。特定の外部 gsQCA パッケージの完全互換クローンとしては
位置づけていません。

統一スキーマからも作成できます:

```python
Expand All @@ -186,8 +190,8 @@ schema = [
{"name": "capacity", "type": "fuzzy", "domain": [0, 1], "calibrated": True},
]

model = MixedQCA.from_schema(
data=mixed_df,
model = GSQCA.from_schema(
data=gsqca_df,
outcome="Y",
schema=schema,
case_id="case",
Expand Down Expand Up @@ -438,7 +442,7 @@ print(result.best_qca_result)
`conservative_models` による理論的制約が必要です。

crisp、fuzzy、multi-valueを混在させる場合は
`fit_mixedqca_from_predictor()` から `MixedQCA` に接続できます。
`fit_gsqca_from_predictor()` から `GSQCA` に接続できます。
fuzzy anchor候補は `run_anchor_sensitivity()` で感度分析できます。

### CV / bootstrap安定性
Expand Down Expand Up @@ -500,7 +504,7 @@ byte単位で同じtreeや順位を生成することを保証するものでは

PyQCA は、`QCAEngineBase` を共通基底として、`CSQCA`, `MVQCA`, `FSQCA` を first-class engine として実装しています。

`MixedQCA` は、それらと同じ共通コアを使う高レベル facade です。crisp / multi-value / fuzzy 条件を同じ分析ワークフローで扱うための入口として設計されています。
`GSQCA` は、それらと同じ共通コアを使う高レベル facade です。crisp / multi-value / fuzzy 条件を同じ分析ワークフローで扱うための入口として設計されています。

```text
qca
Expand All @@ -520,7 +524,7 @@ qca
│ ├── csqca.py
│ ├── fsqca.py
│ ├── mvqca.py
│ └── mixed.py
│ └── gsqca.py
├── minimizers
│ ├── algorithms.py / engine.py / implicant.py / remainder.py
│ ├── backends.py / benchmark.py
Expand All @@ -529,7 +533,7 @@ qca
├── mlqca
│ ├── backend.py / config.py / schema.py / validation.py
│ ├── xgboost.py / importance.py / cutoffs.py
│ ├── calibration.py / csqca.py / fuzzy.py / mixed.py
│ ├── calibration.py / csqca.py / fuzzy.py / gsqca.py
│ ├── search.py / stability.py / results.py
│ └── reporting.py / reproducibility.py / viz.py
├── reporting
Expand Down Expand Up @@ -569,9 +573,9 @@ candidate model、Pareto frontier、CV/bootstrap安定性表を確認できま
QCAとmlQCAをblack boxとして扱わず、予測上の根拠、calibrationの選択、
失敗したmodel評価、resamplingによる変動を追跡可能にします。

### 3. Mixed-typeをfirst-classに扱う
### 3. GSQCAをfirst-classに扱う

crisp、multi-value、fuzzy条件の混在を例外扱いせず、`MixedQCA` の主要な
crisp、multi-value、fuzzy条件の混在を例外扱いせず、`GSQCA` の主要な
利用形態として扱います。

### 4. 最小化アルゴリズムを交換可能にする
Expand Down Expand Up @@ -603,7 +607,7 @@ QCAの方法論的仮定を置き換えるものではありません。
| scpQCA系algorithm | set-coveringを利用する代替的な最小化 |
| ThSQCA | threshold-sweepによる感度分析 |
| R `mlQCA` package | predictive condition selectionとcalibration evidenceを用いる公開mlQCAプロトコル |
| PyQCA | core/mixed QCA、最小化、感度分析、独立実装の`qca.mlqca`を統合するPython-native platform |
| PyQCA | core QCA / GSQCA、最小化、感度分析、独立実装の`qca.mlqca`を統合するPython-native platform |

PyQCAは既存ツールを置き換えるのではなく、相互検証可能なPython実装として
補完することを目指します。`qca.mlqca` はHuangのmlQCA研究を方法論的に
Expand Down
Loading