transport-runtime 0.6.0__tar.gz

transport_runtime-0.6.0/LICENSE

@@ -0,0 +1,25 @@
+MIT License
+
+Copyright (c) 2026 Shengjie Zhang, Wenhao Tian (Shihezi University School of Medicine)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+NOTE: The MIT license covers code only. Data are CDC public-use files
+(NHANES, BRFSS) under their own terms; derived statistical artifacts in
+results/ are released alongside the paper under CC BY 4.0.

transport_runtime-0.6.0/PKG-INFO

@@ -0,0 +1,109 @@
+Metadata-Version: 2.4
+Name: transport-runtime
+Version: 0.6.0
+Summary: Zero-label deployment runtime for fitted classifiers under transport drift
+Author-email: Shengjie Zhang <2142024286@qq.com>, Wenhao Tian <2144767610@qq.com>
+License-Expression: MIT
+Keywords: transportability,calibration,conformal prediction,distribution shift,drift audit
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=2.0
+Requires-Dist: pandas>=1.5
+Requires-Dist: scipy>=1.10
+Provides-Extra: xgb
+Requires-Dist: xgboost>=1.7; extra == "xgb"
+Provides-Extra: test
+Requires-Dist: pytest>=7; extra == "test"
+Requires-Dist: xgboost>=1.7; extra == "test"
+Dynamic: license-file
+
+# Transport Runtime
+
+Zero-label deployment for any fitted probabilistic classifier.
+
+Models are trained on one distribution and executed on another. Transport Runtime wraps any
+fitted classifier with a transport layer: a hot-swappable level interface (exact per-channel
+calibration from public prevalences only), a drift-law audit (three-line report + level
+contamination index), and conformal certificates.
+
+```python
+from transport_runtime import TransportLayer
+
+layer = TransportLayer(model).fit_calibration(X_src, y_src, channels_src)
+layer.audit(X_new, channels_new, y_new)            # three-line drift report + HCI
+deployed = layer.deploy(pi_public, X_unlab, ch)    # zero-label deployment
+deployed.predict_proba(X_new, channels_new)        # per-channel exact level (Prop 1)
+conf = layer.conformal(X_unlab, ch, pi_public, alpha=0.1, per_channel=True)
+conf.predict_set(X_new, channels_new)              # prediction sets with certificates
+```
+
+## Theorem index (each function is one theorem, with citation and validation record)
+
+| Function | Theorem | Source | Validation |
+|---|---|---|---|
+| `moment_delta` / `deploy` | Prop 1: per-channel moment identity — deployed level error identically zero | this project (docs/pilevel_theory.md) | results/pilevel_v08_seeds10.csv, v09 (10 seeds, O/E=1.0000 ×10/10) |
+| `audit` (intercept share, warning) | Drift law: level/slope share decomposition | this project (case series, docs/case_*/DRIFTLAW.md) | six case CSVs in results/ |
+| `hci` | HCI: level-contamination index (slope of per-channel mean margin on logit prevalence) | this project (docs/levelfree_theory.md) | XGB 1.041 vs randomized-training 1.066 (LevelFree falsification record) |
+| `conformal(weighted=True)` | Weighted conformal under covariate shift | Tibshirani et al. 2019 | results/transport_cert_v01.csv (marginal coverage ≥ nominal; per-channel gaps persist) |
+| `conformal(per_channel=True)` | Per-channel thresholds (two-tier: level layer + residual slope share) | this project | results/transport_cert_v02.csv (T3 per-channel coverage 0.896–0.897 flat) |
+| `ot_gap.cdf_l1_gap` | Total coverage gap = ∫p(t)\|F_P−F_Q\|dt (change of variables) + DKW | Correia & Louizos 2025 (arXiv:2507.10425) Thm 3.2/A.2 | synthetic 0.2545 vs 0.2544; T1 real 0.0485 vs empirical 0.0415 |
+| `ot_gap.label_free_gap` | Label-free bound via auxiliary score distributions | Correia & Louizos 2025 Thm 3.3 | validity depends on dominance: `aux_minmax` always valid; `aux_uniform/aux_model` require model accuracy — checked via `check_dominance` |
+| `fit_channel_affine` / `channel_affine_repair` | Prop 4(iii)/(iv): mean-one slope-repair boundary (audited families keep E[M]=1; no universal guarantee) | this project | results/prop4iii_boundary_audit_summary.csv, results/brfss_slope_repair_v01_summary.csv |
+| `decision_curve` / `clinical_threshold_audit` | Net-benefit decision-curve transport evidence | this project | results/*_dca_v01_summary.csv (13 outcome lines) |
+| `transportability_id` / `valid_backdoor_set` / `s_admissible` | Conservative transport-by-adjustment identification: backdoor validity + S-admissibility on explicit DAGs | Pearl–Bareinboim selection-diagram programme; implementation is **not** a complete mz-transportability algorithm | `transport_runtime/tests/test_identify.py` |
+
+## Design principles
+
+1. **Zero-label deployment**: target side needs only unlabeled features + public per-channel prevalences.
+2. **Two-tier certificates mirror the drift law**: level (intercept) share → moment-matched level
+   layer; residual slope share → per-channel conformal thresholds.
+3. **Negative results are documented**: audited training-time level-intervention families fail to remove level
+   content (impossibility duality, docs/theory_impossibility.md, Props 4–5, bounded to audited families) — within those audited families the runtime is the
+   complete solution set, which is why this is a runtime and not a training method.
+4. **Honest bounds**: certificates widen/fail visibly under slope corruption (Prop 2′) rather
+   than failing silently.
+
+## Scope and limits
+
+- Within-channel ranking must transport (Prop 2′: residual ≤ ECE_w + L·τ). If discrimination
+  itself collapses, no level layer can repair it — audit will say so.
+- Small channels: level offsets are shrunk toward a global delta (kappa=200); per-channel
+  conformal degrades in small/rare-event channels (empirical: T2 band 0).
+- The package does not improve iid performance; its ceiling is parity (that is a theorem, not
+  a limitation of implementation).
+
+## Roadmap
+
+- [x] `identify` module: conservative transportability / S-admissibility checker for explicit DAGs (Pearl–Bareinboim adjustment formula; not a complete mz-transportability engine)
+- [ ] OT weight-learning (Correia & Louizos 2025 Sec. 4)
+- [x] transport-suite v1 step A: aggregator over audited `results/*.csv` → `results/transport_suite_v01_summary.csv` and `results/transport_suite_v01_cert_summary.csv`
+- [x] transport-suite v1 step B: low-compute synthetic regression runner (`scripts/transport_suite_v01_runner.py` → `results/transport_suite_v01_runner.csv`)
+- [x] robust layer v0.4: EB uncertainty-aware repair + interval bounds (`transport_runtime.robust`)
+- [x] compiler layer v0.4: TransportCompiler / DeploymentContract with adaptive point-vs-EB repair gating and archive manifest input fingerprints
+- [x] channelization layer v0.4: automatic score-channel discovery from unlabeled covariates and fitted model scores
+- [x] channel selection layer v0.4: transport-aware candidate channel selection balancing level structure and shift risk, with bootstrap stability selection
+- [x] slope-risk + repair layer v0.4: within-channel score-shift / KS diagnosis, limited-label channel-affine repair, validation-based repair-value estimation, and threshold sensitivity analysis
+- [x] slope uncertainty certificate (post-v0.5 working tree): `bootstrap_channel_affine` returns per-channel slope/intercept percentile CIs and success rates; `channel_affine_margin_interval` converts those coefficient CIs into repaired-margin intervals. This quantifies finite-label fitting uncertainty; it is not a proof of target-domain calibration.
+- [x] system demo: end-to-end compiler contract generation with archive manifest, slope repair value estimation, slope repair, certificate replay, and verification (`scripts/ncs_system_v01.py`)
+- [x] verification layer v0.4: contract self-checks for prevalence validity, certificate alpha, slope-risk finiteness, and input fingerprint matching
+- [x] certificate recommendation + replay layer v0.4: risk-aware policy selection and labeled-target coverage evaluation; post-v0.5 `summarize_slope_uncertainty` incorporates finite-label slope-repair CI widths into certificate policy recommendation, `recommend_certificate_from_audit_pack` consumes audit packs directly, and `summarize_channel_clinical_risk` lets recommendations respond to channel-level clinical losses
+- [x] audit pack layer v0.4: standardized JSON artifact combining manifest, slope risk, verification, and certificate evaluation; post-v0.5 `slope_boot=True` adds finite-label slope-repair coefficient CIs and repaired-margin interval summaries, `clinical=True` adds global threshold-specific net-benefit summaries, and `channel_clinical=True` adds per-channel clinical utility summaries
+- [x] clinical decision-curve layer (post-v0.5 working tree): `decision_curve`, `compare_decision_curves`, `clinical_threshold_audit`, and `channel_clinical_audit` evaluate threshold-specific net benefit for unrepaired vs repaired deployment outputs
+- [x] release artifact builder with SHA256 checksums (`scripts/release_v01.py`; current built version `0.5.0`)
+- [x] contract diff layer v0.4: explicit version-to-version comparison for deployment contracts
+- [x] full-system benchmark v0.4: 10-seed clean/level/slope/level+slope scenarios with contract, flagged-only slope repair, and risk-proportional label allocation (`scripts/ncs_fullsystem_v01.py`)
+- [x] figure generation: NCS-oriented benchmark figures and interface schematic (`scripts/ncs_figures_v01.py`, `scripts/ncs_schematic_v01.py` → `results/figs/`)
+- [ ] PyPI release
+
+## Citation
+
+Transport Runtime. Shengjie Zhang, Shihezi University School of Medicine. ORCID 0009-0000-3975-6954.
+Archived at ScienceDB (DOI pending). License: MIT.

transport_runtime-0.6.0/README.md

@@ -0,0 +1,64 @@
+# transport-runtime — zero-label model transport with provable level-repair guarantees
+
+面向 NCS（Nature Computational Science）投稿的计算研究项目：当预测模型部署到新人群、
+新地区或新年代时，零标签校准（zero-label level repair）在什么条件下是**数学保证成立**
+的，什么时候会失败；以及配套的运行时审计协议与开源软件。
+
+## 核心主张（248 臂真实数据证据链支撑，13 结局线，三科学域）
+
+0. **校准-判别分离**：level 修复改善校准（52/64 臂 ECE 降）而判别力冻结（max dAUC 0.009）——fig21/22 系统证据
+1. **Level-drift law**：跨人群/跨周期/跨学科部署时 naive 预测系统漂移（OE 0.39–14.96，政治科学臂创全链纪录），
+   漂移幅度随间隔单调增长（I→P 2.26 → G→I 2.06 → G→P 2.40），
+   跨人群漂移远大于同人群时间漂移（1.21 vs ≤1.08）。
+2. **Level 修复定理**：在满足 (C)（强 log-concave）+ E[M]=1 + 对称单峰的条件下，
+   level repair 使 R(π) ≤ 1 对一切 π 成立（Prop 4(iv) 强形式，第 30 轮解析证明；
+   ψ-Jensen 引理 + Rojo & He 1991 保持定理）。
+3. **Slope 修复边界**：真实数据上 slope 修复无校准收益（ECE 反升）、无临床净获益
+   （DCA 阴性）——conservative 设计（audit + certificate + 最小样本门控）获实证支持。
+
+## 仓库结构
+
+- `transport_runtime/` 可安装 Python 包（审计核心、临床模块、斜率修复、理论证书）
+- `transport_runtime/tests/` 模块测试（沙箱 standalone：84 通过 / 3 可选跳过；含可选共 87 项）
+- `scripts/` 全部确定性脚本（121 个；52 个入 reproduce_all 管线）
+- `results/` 全部产物 CSV/JSON/TEX/PNG（程序读数，无手抄值）
+- `results/figs/fig14_evidence_forest.png` 248 臂证据森林图
+- `results/tables/tab_evidence_summary.tex` 248 行三线表（booktabs 纪律自检；0% 公共患病率误差臂）
+- `preprint/transport_runtime_compendium_v01.md` 合集式大预印本（v0.7.33）
+- `docs/` 理论文档（theory_impossibility.md 第 28–30 轮：保持定理核验与 ψ-Jensen 证明）
+- `manuscript/theorem_index.md` 定理索引
+- `results/ARM_SOURCE_INDEX.csv` 248 臂→源 CSV 可追溯索引
+- `transport_runtime/fetch_data.py` + `transport-runtime-fetch` CLI（PyPI 一键获取全部文件，URL 发布时填入）
+- `docs/AUDIT_CLOSURE_SUMMARY.md` 补漏洞总验收单（41 轮，42+ 攻击面）
+- `docs/PUBLICATION_STRATEGY_DECISIONS.md` 发表策略决策 D1-D5
+- `docs/REFERENCE_MASTER_LIST.md` 统一参考文献主表 + PubMed 核验
+- `docs/RELEASE_ARCHIVAL_CHECKLIST.md` M7 发布清单（ScienceDB-only，Zenodo 禁用）
+- `manuscript/` DRAFT_BUDGET / cover_letter_skeleton / SI_SKELETON / updated_abstract_150w_draft（写作脚手架，prose 待授权）
+- `HANDOFF.md` 跨会话交接卡（新会话必读）
+- `logs/` 轮次日志
+
+## 一键复现
+
+```bash
+PYTHONPATH=. python scripts/reproduce_all_v01.py            # 创建校验清单（全链约 10 分钟，两段式）
+PYTHONPATH=. python scripts/reproduce_all_v01.py --verify   # 验证字节级一致
+```
+
+全部 248 臂证据、机制探针、理论证书、森林图数据、三线表从原始公共数据
+（NHANES 2011-2020 三个周期 + BRFSS 2016-2024）以固定种子精确再生。
+
+## 数据可得性
+
+- NHANES：https://wwwn.cdc.gov/Nchs/Data/Nhanes/Public/ （示范格式
+  `https://wwwn.cdc.gov/Nchs/Data/Nhanes/Public/2021/DataFiles/DEMO_L.xpt`；
+  本项目使用 2011-2012(G)、2015-2016(I)、2017-2020(P) 周期公开文件，本地镜像于 data/）
+- BRFSS：https://www.cdc.gov/brfss/annual_data/annual_YYYY.html （2016-2024 年度文件，本地镜像于 data/brfss/）
+
+## 作者
+
+- 共同第一：Shengjie Zhang（石河子大学医学院，ORCID 0009-0000-3975-6954）
+- 共同第一兼通讯：Wenhao Tian（石河子大学医学院，ORCID 0009-0002-0660-3024）
+
+## 许可
+
+代码 MIT；数据为 CDC 公开数据，遵循各原始使用条款。

transport_runtime-0.6.0/pyproject.toml

@@ -0,0 +1,47 @@
+[build-system]
+requires = ["setuptools>=61", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "transport-runtime"
+version = "0.6.0"
+description = "Zero-label deployment runtime for fitted classifiers under transport drift"
+readme = "transport_runtime/README.md"
+requires-python = ">=3.10"
+license = "MIT"
+license-files = ["LICENSE"]
+authors = [
+  {name = "Shengjie Zhang", email = "2142024286@qq.com"},
+  {name = "Wenhao Tian", email = "2144767610@qq.com"}
+]
+keywords = ["transportability", "calibration", "conformal prediction", "distribution shift", "drift audit"]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "Intended Audience :: Science/Research",
+  "Programming Language :: Python :: 3",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Topic :: Scientific/Engineering"
+]
+dependencies = [
+  "numpy>=2.0",
+  "pandas>=1.5",
+  "scipy>=1.10"
+]
+
+[project.optional-dependencies]
+xgb = ["xgboost>=1.7"]
+test = ["pytest>=7", "xgboost>=1.7"]
+
+[project.scripts]
+transport-runtime = "transport_runtime.cli:main"
+transport-runtime-fetch = "transport_runtime.fetch_data:main"
+
+[tool.setuptools]
+packages = ["transport_runtime"]
+package-dir = {"" = "."}
+include-package-data = false
+
+[tool.pytest.ini_options]
+testpaths = ["transport_runtime/tests"]

transport_runtime-0.6.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

transport_runtime-0.6.0/transport_runtime/README.md

@@ -0,0 +1,82 @@
+# Transport Runtime
+
+Zero-label deployment for any fitted probabilistic classifier.
+
+Models are trained on one distribution and executed on another. Transport Runtime wraps any
+fitted classifier with a transport layer: a hot-swappable level interface (exact per-channel
+calibration from public prevalences only), a drift-law audit (three-line report + level
+contamination index), and conformal certificates.
+
+```python
+from transport_runtime import TransportLayer
+
+layer = TransportLayer(model).fit_calibration(X_src, y_src, channels_src)
+layer.audit(X_new, channels_new, y_new)            # three-line drift report + HCI
+deployed = layer.deploy(pi_public, X_unlab, ch)    # zero-label deployment
+deployed.predict_proba(X_new, channels_new)        # per-channel exact level (Prop 1)
+conf = layer.conformal(X_unlab, ch, pi_public, alpha=0.1, per_channel=True)
+conf.predict_set(X_new, channels_new)              # prediction sets with certificates
+```
+
+## Theorem index (each function is one theorem, with citation and validation record)
+
+| Function | Theorem | Source | Validation |
+|---|---|---|---|
+| `moment_delta` / `deploy` | Prop 1: per-channel moment identity — deployed level error identically zero | this project (docs/pilevel_theory.md) | results/pilevel_v08_seeds10.csv, v09 (10 seeds, O/E=1.0000 ×10/10) |
+| `audit` (intercept share, warning) | Drift law: level/slope share decomposition | this project (case series, docs/case_*/DRIFTLAW.md) | six case CSVs in results/ |
+| `hci` | HCI: level-contamination index (slope of per-channel mean margin on logit prevalence) | this project (docs/levelfree_theory.md) | XGB 1.041 vs randomized-training 1.066 (LevelFree falsification record) |
+| `conformal(weighted=True)` | Weighted conformal under covariate shift | Tibshirani et al. 2019 | results/transport_cert_v01.csv (marginal coverage ≥ nominal; per-channel gaps persist) |
+| `conformal(per_channel=True)` | Per-channel thresholds (two-tier: level layer + residual slope share) | this project | results/transport_cert_v02.csv (T3 per-channel coverage 0.896–0.897 flat) |
+| `ot_gap.cdf_l1_gap` | Total coverage gap = ∫p(t)\|F_P−F_Q\|dt (change of variables) + DKW | Correia & Louizos 2025 (arXiv:2507.10425) Thm 3.2/A.2 | synthetic 0.2545 vs 0.2544; T1 real 0.0485 vs empirical 0.0415 |
+| `ot_gap.label_free_gap` | Label-free bound via auxiliary score distributions | Correia & Louizos 2025 Thm 3.3 | validity depends on dominance: `aux_minmax` always valid; `aux_uniform/aux_model` require model accuracy — checked via `check_dominance` |
+| `fit_channel_affine` / `channel_affine_repair` | Prop 4(iii)/(iv): mean-one slope-repair boundary (audited families keep E[M]=1; no universal guarantee) | this project | results/prop4iii_boundary_audit_summary.csv, results/brfss_slope_repair_v01_summary.csv |
+| `decision_curve` / `clinical_threshold_audit` | Net-benefit decision-curve transport evidence | this project | results/*_dca_v01_summary.csv (13 outcome lines) |
+| `transportability_id` / `valid_backdoor_set` / `s_admissible` | Conservative transport-by-adjustment identification: backdoor validity + S-admissibility on explicit DAGs | Pearl–Bareinboim selection-diagram programme; implementation is **not** a complete mz-transportability algorithm | `transport_runtime/tests/test_identify.py` |
+
+## Design principles
+
+1. **Zero-label deployment**: target side needs only unlabeled features + public per-channel prevalences.
+2. **Two-tier certificates mirror the drift law**: level (intercept) share → moment-matched level
+   layer; residual slope share → per-channel conformal thresholds.
+3. **Negative results are documented**: audited training-time level-intervention families fail to remove level
+   content (impossibility duality, docs/theory_impossibility.md, Props 4–5, bounded to audited families) — within those audited families the runtime is the
+   complete solution set, which is why this is a runtime and not a training method.
+4. **Honest bounds**: certificates widen/fail visibly under slope corruption (Prop 2′) rather
+   than failing silently.
+
+## Scope and limits
+
+- Within-channel ranking must transport (Prop 2′: residual ≤ ECE_w + L·τ). If discrimination
+  itself collapses, no level layer can repair it — audit will say so.
+- Small channels: level offsets are shrunk toward a global delta (kappa=200); per-channel
+  conformal degrades in small/rare-event channels (empirical: T2 band 0).
+- The package does not improve iid performance; its ceiling is parity (that is a theorem, not
+  a limitation of implementation).
+
+## Roadmap
+
+- [x] `identify` module: conservative transportability / S-admissibility checker for explicit DAGs (Pearl–Bareinboim adjustment formula; not a complete mz-transportability engine)
+- [ ] OT weight-learning (Correia & Louizos 2025 Sec. 4)
+- [x] transport-suite v1 step A: aggregator over audited `results/*.csv` → `results/transport_suite_v01_summary.csv` and `results/transport_suite_v01_cert_summary.csv`
+- [x] transport-suite v1 step B: low-compute synthetic regression runner (`scripts/transport_suite_v01_runner.py` → `results/transport_suite_v01_runner.csv`)
+- [x] robust layer v0.4: EB uncertainty-aware repair + interval bounds (`transport_runtime.robust`)
+- [x] compiler layer v0.4: TransportCompiler / DeploymentContract with adaptive point-vs-EB repair gating and archive manifest input fingerprints
+- [x] channelization layer v0.4: automatic score-channel discovery from unlabeled covariates and fitted model scores
+- [x] channel selection layer v0.4: transport-aware candidate channel selection balancing level structure and shift risk, with bootstrap stability selection
+- [x] slope-risk + repair layer v0.4: within-channel score-shift / KS diagnosis, limited-label channel-affine repair, validation-based repair-value estimation, and threshold sensitivity analysis
+- [x] slope uncertainty certificate (post-v0.5 working tree): `bootstrap_channel_affine` returns per-channel slope/intercept percentile CIs and success rates; `channel_affine_margin_interval` converts those coefficient CIs into repaired-margin intervals. This quantifies finite-label fitting uncertainty; it is not a proof of target-domain calibration.
+- [x] system demo: end-to-end compiler contract generation with archive manifest, slope repair value estimation, slope repair, certificate replay, and verification (`scripts/ncs_system_v01.py`)
+- [x] verification layer v0.4: contract self-checks for prevalence validity, certificate alpha, slope-risk finiteness, and input fingerprint matching
+- [x] certificate recommendation + replay layer v0.4: risk-aware policy selection and labeled-target coverage evaluation; post-v0.5 `summarize_slope_uncertainty` incorporates finite-label slope-repair CI widths into certificate policy recommendation, `recommend_certificate_from_audit_pack` consumes audit packs directly, and `summarize_channel_clinical_risk` lets recommendations respond to channel-level clinical losses
+- [x] audit pack layer v0.4: standardized JSON artifact combining manifest, slope risk, verification, and certificate evaluation; post-v0.5 `slope_boot=True` adds finite-label slope-repair coefficient CIs and repaired-margin interval summaries, `clinical=True` adds global threshold-specific net-benefit summaries, and `channel_clinical=True` adds per-channel clinical utility summaries
+- [x] clinical decision-curve layer (post-v0.5 working tree): `decision_curve`, `compare_decision_curves`, `clinical_threshold_audit`, and `channel_clinical_audit` evaluate threshold-specific net benefit for unrepaired vs repaired deployment outputs
+- [x] release artifact builder with SHA256 checksums (`scripts/release_v01.py`; current built version `0.5.0`)
+- [x] contract diff layer v0.4: explicit version-to-version comparison for deployment contracts
+- [x] full-system benchmark v0.4: 10-seed clean/level/slope/level+slope scenarios with contract, flagged-only slope repair, and risk-proportional label allocation (`scripts/ncs_fullsystem_v01.py`)
+- [x] figure generation: NCS-oriented benchmark figures and interface schematic (`scripts/ncs_figures_v01.py`, `scripts/ncs_schematic_v01.py` → `results/figs/`)
+- [ ] PyPI release
+
+## Citation
+
+Transport Runtime. Shengjie Zhang, Shihezi University School of Medicine. ORCID 0009-0000-3975-6954.
+Archived at ScienceDB (DOI pending). License: MIT.

transport_runtime-0.6.0/transport_runtime/__init__.py

@@ -0,0 +1,85 @@
+"""Transport Runtime: zero-label deployment for any fitted classifier."""
+from .core import TransportLayer, DeployedModel, DeployedConformal, moment_delta
+from .ot_gap import (cdf_l1_gap, label_free_gap, aux_minmax, aux_uniform,
+                     aux_model, total_coverage_gap)
+from .channels import discover_score_channels, quantile_channel
+from .clinical import channel_clinical_audit, clinical_threshold_audit, compare_decision_curves, decision_curve
+from .selection import select_transport_channel, select_transport_channel_stability
+from .slope import (
+    allocate_labels_for_slope,
+    apply_channel_affine,
+    channel_affine_repair,
+    channel_affine_margin_interval,
+    bootstrap_channel_affine,
+    ece_from_p,
+    evaluate_affine_repair,
+    fit_channel_affine,
+    margins_from_model,
+    slope_threshold_sensitivity,
+    summarize_slope_risk,
+    within_channel_score_shift,
+)
+from .certify import CertificateRecommendation, evaluate_certificate, recommend_certificate, recommend_certificate_from_audit_pack, summarize_channel_clinical_risk, summarize_slope_uncertainty
+from .audit import audit_pack, save_audit_pack
+from .diff import contract_diff
+from .verify import CheckResult, VerificationReport, verify_contract
+from .compiler import (
+    CertificateSpec,
+    ChannelSpec,
+    DeploymentContract,
+    PrevalenceSpec,
+    TransportCompiler,
+)
+from .robust import (
+    PrevalenceIntervalSet,
+    choose_repair,
+    deploy_eb,
+    deploy_interval,
+    empirical_bayes_target,
+    estimate_repair_gain,
+    interval_regret_bound,
+    interval_uniform_variance,
+    minimax_target,
+    worst_case_channel_bias,
+)
+from .identify import (
+    IdentificationResult,
+    STATUS_INCONCLUSIVE,
+    STATUS_NOT_TRANSPORTABLE_BY_ADJUSTMENT,
+    STATUS_TRANSPORTABLE,
+    ancestors,
+    d_connected,
+    descendants,
+    intervened_graph,
+    is_dag,
+    s_admissible,
+    transportability_id,
+    valid_backdoor_set,
+)
+
+__all__ = [
+    'TransportLayer', 'DeployedModel', 'DeployedConformal', 'moment_delta',
+    'ChannelSpec', 'PrevalenceSpec', 'CertificateSpec', 'TransportCompiler', 'DeploymentContract',
+    'quantile_channel', 'discover_score_channels', 'select_transport_channel', 'select_transport_channel_stability',
+    'decision_curve', 'compare_decision_curves', 'clinical_threshold_audit', 'channel_clinical_audit',
+    'margins_from_model', 'within_channel_score_shift', 'summarize_slope_risk', 'slope_threshold_sensitivity',
+    'fit_channel_affine', 'apply_channel_affine', 'channel_affine_repair',
+    'bootstrap_channel_affine', 'channel_affine_margin_interval',
+    'allocate_labels_for_slope',
+    'ece_from_p', 'evaluate_affine_repair',
+    'verify_contract', 'VerificationReport', 'CheckResult',
+    'recommend_certificate', 'recommend_certificate_from_audit_pack', 'CertificateRecommendation', 'evaluate_certificate', 'summarize_slope_uncertainty', 'summarize_channel_clinical_risk',
+    'audit_pack', 'save_audit_pack', 'contract_diff',
+    'cdf_l1_gap', 'label_free_gap', 'aux_minmax', 'aux_uniform',
+    'aux_model', 'total_coverage_gap',
+    'PrevalenceIntervalSet', 'deploy_interval', 'interval_regret_bound',
+    'deploy_eb', 'empirical_bayes_target', 'estimate_repair_gain', 'choose_repair', 'interval_uniform_variance',
+    'minimax_target', 'worst_case_channel_bias',
+    'IdentificationResult', 'transportability_id', 'valid_backdoor_set',
+    's_admissible', 'd_connected', 'intervened_graph', 'is_dag',
+    'ancestors', 'descendants', 'STATUS_TRANSPORTABLE', 'STATUS_INCONCLUSIVE',
+    'STATUS_NOT_TRANSPORTABLE_BY_ADJUSTMENT',
+]
+__version__ = '0.6.0'
+
+from transport_runtime.pipeline import TransportPipeline  # noqa: E402,F401

transport_runtime-0.6.0/transport_runtime/audit.py

@@ -0,0 +1,210 @@
+"""Standardized audit pack for deployment contracts.
+
+An audit pack is the single JSON artifact that combines:
+- contract manifest;
+- slope-risk summary;
+- contract verification;
+- labeled-target certificate replay, when labels are available.
+"""
+from __future__ import annotations
+
+import json
+
+import numpy as np
+from scipy.special import expit
+
+from .certify import evaluate_certificate
+from .clinical import clinical_threshold_audit, decision_curve
+from .compiler import DeploymentContract
+from .slope import bootstrap_channel_affine, channel_affine_margin_interval, margins_from_model
+from .verify import verify_contract
+
+
+def _tolist(value):
+    return np.asarray(value, dtype=float).tolist()
+
+
+def _slope_repair_certificate(contract, X, channels, y, boot_kwargs):
+    levels = getattr(contract.deployed, "levels_", None)
+    if levels is None:
+        return None
+    f_repaired = margins_from_model(contract.layer.model, X) + np.asarray(levels, dtype=float)[np.asarray(channels, dtype=int)]
+    boot = bootstrap_channel_affine(f_repaired, y, channels, **boot_kwargs)
+    interval = channel_affine_margin_interval(f_repaired, channels, boot)
+    width = interval["upper"] - interval["lower"]
+    return {
+        "note": "finite-label uncertainty for affine slope repair fitted after level repair; "
+                "this is fitting uncertainty, not proof of target calibration",
+        "alpha": float(boot["alpha"]),
+        "ci_level": float(boot["ci_level"]),
+        "n_boot": int(boot["n_boot"]),
+        "channel_n": _tolist(boot["channel_n"]),
+        "a": _tolist(boot["a"]),
+        "b": _tolist(boot["b"]),
+        "a_mean": _tolist(boot["a_mean"]),
+        "b_mean": _tolist(boot["b_mean"]),
+        "a_std": _tolist(boot["a_std"]),
+        "b_std": _tolist(boot["b_std"]),
+        "a_ci_lower": _tolist(boot["a_ci_lower"]),
+        "a_ci_upper": _tolist(boot["a_ci_upper"]),
+        "b_ci_lower": _tolist(boot["b_ci_lower"]),
+        "b_ci_upper": _tolist(boot["b_ci_upper"]),
+        "bootstrap_success_rate": _tolist(boot["bootstrap_success_rate"]),
+        "margin_interval_summary": {
+            "n": int(len(width)),
+            "mean_width": float(np.mean(width)),
+            "median_width": float(np.median(width)),
+            "q95_width": float(np.quantile(width, 0.95)),
+            "max_width": float(np.max(width)),
+        },
+    }
+
+
+def _clinical_threshold_audit(contract, X, channels, y, thresholds):
+    base_p = expit(margins_from_model(contract.layer.model, X))
+    repaired_p = contract.predict_proba(X, channels)
+    audit = clinical_threshold_audit(y, base_p, repaired_p, thresholds=thresholds)
+    return {
+        "note": "threshold-specific net-benefit audit comparing repaired contract output with unrepaired model output; "
+                "synthetic or labeled-target audit evidence does not by itself prove clinical effectiveness",
+        "thresholds": audit["thresholds"],
+        "base_net_benefit": audit["base_net_benefit"],
+        "repaired_net_benefit": audit["repaired_net_benefit"],
+        "repaired_minus_base": audit["repaired_minus_base"],
+        "summary": audit["summary"],
+    }
+
+
+def _channel_clinical_summary(contract, X, channels, y, thresholds):
+    thresholds = np.asarray(
+        thresholds if thresholds is not None else np.linspace(0.01, 0.99, 99),
+        dtype=float,
+    )
+    base_p = expit(margins_from_model(contract.layer.model, X))
+    repaired_p = contract.predict_proba(X, channels)
+    y = np.asarray(y, dtype=float)
+    channels = np.asarray(channels, dtype=int)
+    summaries = []
+    for g in np.unique(channels):
+        sel = channels == g
+        base = decision_curve(y[sel], base_p[sel], thresholds=thresholds)
+        repaired = decision_curve(y[sel], repaired_p[sel], thresholds=thresholds)
+        gain = np.asarray(repaired["net_benefit"], dtype=float) - np.asarray(base["net_benefit"], dtype=float)
+        j = int(np.argmax(gain))
+        summaries.append(
+            {
+                "channel": int(g),
+                "channel_n": int(sel.sum()),
+                "channel_prevalence": float(np.mean(y[sel])),
+                "integrated_gain": float(np.trapz(gain, thresholds)),
+                "max_gain": float(gain[j]),
+                "threshold_at_max_gain": float(thresholds[j]),
+                "worst_gain": float(np.min(gain)),
+                "fraction_improved": float(np.mean(gain > 0.0)),
+            }
+        )
+    integrated = np.array([row["integrated_gain"] for row in summaries], dtype=float)
+    fragile = summaries[int(np.argmin(integrated))]
+    return {
+        "note": "channel-level threshold-specific net-benefit summary; negative channel gains can be masked by aggregate averages",
+        "thresholds": thresholds.tolist(),
+        "channel_summaries": summaries,
+        "overall": {
+            "n_channels": int(len(summaries)),
+            "mean_integrated_gain": float(np.mean(integrated)),
+            "min_integrated_gain": float(np.min(integrated)),
+            "n_channels_negative_integrated_gain": int(np.sum(integrated < 0.0)),
+            "most_fragile_channel": fragile["channel"],
+            "most_fragile_integrated_gain": fragile["integrated_gain"],
+            "most_fragile_worst_gain": fragile["worst_gain"],
+        },
+    }
+
+
+def audit_pack(
+    contract: DeploymentContract,
+    X=None,
+    channels=None,
+    y=None,
+    slope_boot=False,
+    slope_boot_kwargs=None,
+    clinical=False,
+    clinical_thresholds=None,
+    channel_clinical=False,
+    channel_clinical_thresholds=None,
+) -> dict:
+    """Build a standardized audit pack for a deployment contract.
+
+    Set `slope_boot=True` to add a finite-label affine slope-repair uncertainty
+    certificate when labeled target data are supplied. The option is disabled by
+    default to keep routine audit packs lightweight.
+    """
+    pack = contract.manifest()
+    pack["slope_risk"] = contract.slope_risk
+
+    if X is not None and channels is not None:
+        pack["verification"] = verify_contract(contract, X=X, channels=channels).to_dict()
+    else:
+        pack["verification"] = None
+
+    if contract.certificate is not None and X is not None and y is not None and channels is not None:
+        pack["certificate_evaluation"] = evaluate_certificate(contract, X, y, channels)
+    else:
+        pack["certificate_evaluation"] = None
+
+    if slope_boot:
+        if X is None or y is None or channels is None:
+            raise ValueError("slope_boot=True requires X, channels, and y")
+        kwargs = dict(slope_boot_kwargs or {})
+        pack["slope_repair_certificate"] = _slope_repair_certificate(contract, X, channels, y, kwargs)
+    else:
+        pack["slope_repair_certificate"] = None
+
+    if clinical:
+        if X is None or y is None or channels is None:
+            raise ValueError("clinical=True requires X, channels, and y")
+        pack["clinical_threshold_audit"] = _clinical_threshold_audit(contract, X, channels, y, clinical_thresholds)
+    else:
+        pack["clinical_threshold_audit"] = None
+
+    if channel_clinical:
+        if X is None or y is None or channels is None:
+            raise ValueError("channel_clinical=True requires X, channels, and y")
+        pack["channel_clinical_summary"] = _channel_clinical_summary(
+            contract, X, channels, y, channel_clinical_thresholds
+        )
+    else:
+        pack["channel_clinical_summary"] = None
+
+    return pack
+
+
+def save_audit_pack(
+    contract: DeploymentContract,
+    path,
+    X=None,
+    channels=None,
+    y=None,
+    slope_boot=False,
+    slope_boot_kwargs=None,
+    clinical=False,
+    clinical_thresholds=None,
+    channel_clinical=False,
+    channel_clinical_thresholds=None,
+) -> dict:
+    """Save an audit pack JSON and return it."""
+    pack = audit_pack(
+        contract,
+        X=X,
+        channels=channels,
+        y=y,
+        slope_boot=slope_boot,
+        slope_boot_kwargs=slope_boot_kwargs,
+        clinical=clinical,
+        clinical_thresholds=clinical_thresholds,
+        channel_clinical=channel_clinical,
+        channel_clinical_thresholds=channel_clinical_thresholds,
+    )
+    with open(str(path), "w", encoding="utf-8") as f:
+        json.dump(pack, f, indent=2, sort_keys=True)
+    return pack