| Name | fairodds-auc JSON |
| Version |
0.2.0
JSON |
| download |
| home_page | None |
| Summary | FairOdds-AUC: A fairness-scaled AUROC metric that penalizes equalized odds gaps across sensitive attributes |
| upload_time | 2025-08-20 21:59:37 |
| maintainer | None |
| docs_url | None |
| author | FairOdds-AUC Authors |
| requires_python | >=3.8 |
| license | MIT License
Copyright (c) 2025 FairOdds-AUC Authors
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. |
| keywords |
auc
auroc
bias
equalized odds
fairness
metrics
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
### FairOdds-AUC
FairOdds-AUC is a fairness-scaled AUROC metric that multiplicatively penalizes equalized-odds gaps across user-specified sensitive attributes. It enables a single, tunable score that balances utility and fairness via a nonnegative temperature parameter λ.
Formula:
- FairOdds-AUC = AUROC / (1 + λ · GEO)
- GEO = sum over sensitive attributes of Equalized Odds differences (per Fairlearn)
When λ = 0, the score reduces to AUROC. Larger λ puts more weight on fairness, discounting models with larger equalized-odds disparities.
### Installation
```bash
pip install fairodds-auc
```
From source (editable):
```bash
pip install -U pip
pip install -e .
```
### Quickstart
```python
import numpy as np
from fairodds_auc import fair_odds_auc
y_true = np.array([0, 1, 0, 1, 0, 1])
y_score = np.array([0.1, 0.9, 0.3, 0.8, 0.2, 0.7])
# any attributes you care about
race = np.array(["A", "A", "B", "B", "A", "B"])
sex = np.array(["F", "M", "F", "M", "F", "M"])
sensitive_features = {"race": race, "sex": sex}
score = fair_odds_auc(
y_true=y_true,
y_score=y_score,
sensitive_features=sensitive_features,
lambda_=1.0,
threshold=0.5,
agg="mean", # or 'worst_case' to match Fairlearn default
)
print(score)
```
### API
- `fair_odds_auc(y_true, y_score, sensitive_features, lambda_=1.0, threshold=0.5, method='between_groups', agg='mean', sample_weight=None) -> float`
- `equalized_odds_gap(y_true, y_pred, group, method='between_groups', agg='mean', sample_weight=None) -> float`
- `generalized_equalized_odds(y_true, y_pred, sensitive_features, method='between_groups', agg='mean', sample_weight=None) -> (dict, float)`
Notes:
- EO uses hard decisions `y_pred` from thresholding `y_score`.
- EO is computed using `fairlearn.metrics.equalized_odds_difference` under the hood.
- Pass `sensitive_features` as a single array-like or a dict of name->array-like.
### References
- Fong et al. (2022): Bias-penalized AUC (fairAUC)
- Pfohl et al. (2021): Balancing performance and fairness in clinical ML
- Dehdashtian et al. (2024): U-FaTE multi-objective framework
### License
MIT
Raw data
{
"_id": null,
"home_page": null,
"name": "fairodds-auc",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": "auc, auroc, bias, equalized odds, fairness, metrics",
"author": "FairOdds-AUC Authors",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/68/b0/8d9ba249724be20b7e69fc660e80c2e224e7d7c2e4611811d9130e1e5fb4/fairodds_auc-0.2.0.tar.gz",
"platform": null,
"description": "### FairOdds-AUC\n\nFairOdds-AUC is a fairness-scaled AUROC metric that multiplicatively penalizes equalized-odds gaps across user-specified sensitive attributes. It enables a single, tunable score that balances utility and fairness via a nonnegative temperature parameter \u03bb.\n\nFormula:\n- FairOdds-AUC = AUROC / (1 + \u03bb \u00b7 GEO)\n- GEO = sum over sensitive attributes of Equalized Odds differences (per Fairlearn)\n\nWhen \u03bb = 0, the score reduces to AUROC. Larger \u03bb puts more weight on fairness, discounting models with larger equalized-odds disparities.\n\n### Installation\n\n```bash\npip install fairodds-auc\n```\n\nFrom source (editable):\n\n```bash\npip install -U pip\npip install -e .\n```\n\n### Quickstart\n\n```python\nimport numpy as np\nfrom fairodds_auc import fair_odds_auc\n\ny_true = np.array([0, 1, 0, 1, 0, 1])\ny_score = np.array([0.1, 0.9, 0.3, 0.8, 0.2, 0.7])\n\n# any attributes you care about\nrace = np.array([\"A\", \"A\", \"B\", \"B\", \"A\", \"B\"]) \nsex = np.array([\"F\", \"M\", \"F\", \"M\", \"F\", \"M\"]) \n\nsensitive_features = {\"race\": race, \"sex\": sex}\n\nscore = fair_odds_auc(\n y_true=y_true,\n y_score=y_score,\n sensitive_features=sensitive_features,\n lambda_=1.0,\n threshold=0.5,\n agg=\"mean\", # or 'worst_case' to match Fairlearn default\n)\nprint(score)\n```\n\n### API\n\n- `fair_odds_auc(y_true, y_score, sensitive_features, lambda_=1.0, threshold=0.5, method='between_groups', agg='mean', sample_weight=None) -> float`\n- `equalized_odds_gap(y_true, y_pred, group, method='between_groups', agg='mean', sample_weight=None) -> float`\n- `generalized_equalized_odds(y_true, y_pred, sensitive_features, method='between_groups', agg='mean', sample_weight=None) -> (dict, float)`\n\nNotes:\n- EO uses hard decisions `y_pred` from thresholding `y_score`.\n- EO is computed using `fairlearn.metrics.equalized_odds_difference` under the hood.\n- Pass `sensitive_features` as a single array-like or a dict of name->array-like.\n\n### References\n- Fong et al. (2022): Bias-penalized AUC (fairAUC)\n- Pfohl et al. (2021): Balancing performance and fairness in clinical ML\n- Dehdashtian et al. (2024): U-FaTE multi-objective framework\n\n### License\nMIT\n",
"bugtrack_url": null,
"license": "MIT License\n \n Copyright (c) 2025 FairOdds-AUC Authors\n \n Permission is hereby granted, free of charge, to any person obtaining a copy\n of this software and associated documentation files (the \"Software\"), to deal\n in the Software without restriction, including without limitation the rights\n to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n copies of the Software, and to permit persons to whom the Software is\n furnished to do so, subject to the following conditions:\n \n The above copyright notice and this permission notice shall be included in all\n copies or substantial portions of the Software.\n \n THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\n SOFTWARE.",
"summary": "FairOdds-AUC: A fairness-scaled AUROC metric that penalizes equalized odds gaps across sensitive attributes",
"version": "0.2.0",
"project_urls": {
"Homepage": "https://github.com/your-name/fairodds-auc",
"Issues": "https://github.com/your-name/fairodds-auc/issues",
"Repository": "https://github.com/your-name/fairodds-auc"
},
"split_keywords": [
"auc",
" auroc",
" bias",
" equalized odds",
" fairness",
" metrics"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "ab86786802c6ea0f07e7cf90321c336b5459ce0c987196354918d713f8d179ca",
"md5": "8d985e2664b895d2d86782837bde2336",
"sha256": "3b678f9eb933ffcc1315091059a19b1f118e78ffe4d3ecc8f03fb6e4f9dd7eac"
},
"downloads": -1,
"filename": "fairodds_auc-0.2.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "8d985e2664b895d2d86782837bde2336",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 7962,
"upload_time": "2025-08-20T21:59:36",
"upload_time_iso_8601": "2025-08-20T21:59:36.912705Z",
"url": "https://files.pythonhosted.org/packages/ab/86/786802c6ea0f07e7cf90321c336b5459ce0c987196354918d713f8d179ca/fairodds_auc-0.2.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "68b08d9ba249724be20b7e69fc660e80c2e224e7d7c2e4611811d9130e1e5fb4",
"md5": "f4504fbdb6fedd4d05a1bcb1873801d9",
"sha256": "5db38b2a9628f41fb6489c79e33f6f301ab2e4cdbb951fbd442d2a8022ce673e"
},
"downloads": -1,
"filename": "fairodds_auc-0.2.0.tar.gz",
"has_sig": false,
"md5_digest": "f4504fbdb6fedd4d05a1bcb1873801d9",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 4379,
"upload_time": "2025-08-20T21:59:37",
"upload_time_iso_8601": "2025-08-20T21:59:37.854634Z",
"url": "https://files.pythonhosted.org/packages/68/b0/8d9ba249724be20b7e69fc660e80c2e224e7d7c2e4611811d9130e1e5fb4/fairodds_auc-0.2.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-20 21:59:37",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "your-name",
"github_project": "fairodds-auc",
"github_not_found": true,
"lcname": "fairodds-auc"
}
