l0l2learn 0.1.0__tar.gz

l0l2learn-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 ml-lab-htw
+
+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.

l0l2learn-0.1.0/PKG-INFO

@@ -0,0 +1,213 @@
+Metadata-Version: 2.4
+Name: l0l2learn
+Version: 0.1.0
+Summary: Cardinality- and budget-constrained feature selection for logistic regression using mixed-integer conic optimization
+Home-page: https://github.com/ml-lab-htw/l0l2learn
+Author: Ricardo Knauer
+Author-email: ricardo.knauer@htw-berlin.de
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: joblib
+Requires-Dist: mosek
+Requires-Dist: numpy
+Requires-Dist: pandas
+Requires-Dist: scikit-learn
+Requires-Dist: scipy
+Requires-Dist: tqdm
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: license-file
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# l0l2learn
+
+Feature selection for logistic regression using mixed-integer conic optimization. Unlike Lasso-based approaches, `l0l2learn` directly optimizes feature subsets under explicit cardinality or budget constraints.
+
+## Overview
+
+`l0l2learn` is a Python package that provides sklearn-style estimators for cardinality- and budget-constrained feature selection in logistic regression. The package currently includes:
+
+- **L0L2Classifier**: L0-constrained L2-regularized logistic regression
+- **ResampledL0L2Classifier**: resampling-based feature selection with frequency-based aggregation to improve the selection stability
+
+## Installation
+
+To install the package, use the following command:
+
+```sh
+pip install l0l2learn
+```
+
+Please check the [MOSEK website](https://www.mosek.com/) to request and set up a license for the conic solver.
+
+## Quick Start
+
+### Feature Selection Without Resampling
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import L0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = L0L2Classifier(
+    b=3,
+    lambd=1.0
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+```
+
+### Feature Selection With Resampling
+
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import ResampledL0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = ResampledL0L2Classifier(
+    b=3,
+    param_grid={"lambd": [1.0]},
+    n_resamples=3
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+print("VIFs:         ", clf.variable_inclusion_frequencies_[clf.support_])
+print("MSFs:         ", clf.model_selection_frequencies_)
+```
+
+## Hyperparameters
+
+### Feature Costs
+
+Feature-specific costs can be supplied through `c`:
+
+```python
+clf = L0L2Classifier(c=[1, 2, 5])
+```
+
+The optimization then accounts for some variables to be more expensive than others.
+
+### Feature Selection Budget
+
+The feature selection budget is controlled through `b`:
+
+```python
+clf = L0L2Classifier(b=5)
+```
+
+When all feature costs are equal to one (the default), `b` directly controls the maximum number of selected features.
+
+### L2 Regularization
+
+The L2 regularization strength is given by `lambd`:
+
+```python
+clf = L0L2Classifier(lambd=0.1)
+```
+
+Larger values can attenuate overfitting and increase robustness.
+
+### Number of Resamples
+
+`n_resamples` determines how many resampled models are fitted:
+
+```python
+clf = ResampledL0L2Classifier(b=5, n_resamples=99)
+```
+
+Larger values can improve frequency estimates but increase runtime.
+
+### Other Hyperparameters
+
+#### L0L2Classifier
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`time_limit`**: Maximum runtime in seconds for the optimization problem.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+#### ResampledL0L2Classifier
+
+- **`resampling`**: Controls whether and how rows, columns, both, or neither are resampled.
+
+- **`n_row_subsamples`**: Number or fraction of observations used during row subsampling.
+
+- **`n_column_subsamples`**: Number or fraction of features used during column subsampling.
+
+- **`aggregation`**: Whether model selection or variable inclusion frequencies are used for aggregation.
+
+- **`vif_threshold`**: Minimum variable inclusion frequency required when using `aggregation="VIF"`.
+
+- **`estimator`**: Alternative base estimator used instead of the default `L0L2Classifier`.
+
+- **`param_grid`**: Hyperparameter grid used for cross-validation when tuning `lambd`.
+
+- **`cv`**: Cross-validation strategy used for hyperparameter tuning.
+
+- **`scoring`**: Scoring metric used to select the best hyperparameter configuration.
+
+- **`numerical_features`**: Specifies which DataFrame columns should be treated as numerical features.
+
+- **`categorical_features`**: Specifies which DataFrame columns should be treated as categorical features.
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`mosek_time_limit`**: Maximum runtime in seconds for each individual optimization problem.
+
+- **`total_time_limit`**: Maximum runtime in seconds for the complete resampling procedure.
+
+- **`max_consecutive_failures`**: Stops resampling if too many consecutive model fits fail.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+- **`n_jobs`**: Number of parallel workers used during resampling.
+
+- **`random_state`**: Controls the randomness of resampling and cross-validation procedures.
+
+## License
+
+This project is licensed under the [MIT License](https://github.com/ml-lab-htw/l0l2learn/blob/main/LICENSE). See the `LICENSE` file for details.
+
+## Authors
+
+- Ricardo Knauer (HTW Berlin)

l0l2learn-0.1.0/README.md

@@ -0,0 +1,182 @@
+# l0l2learn
+
+Feature selection for logistic regression using mixed-integer conic optimization. Unlike Lasso-based approaches, `l0l2learn` directly optimizes feature subsets under explicit cardinality or budget constraints.
+
+## Overview
+
+`l0l2learn` is a Python package that provides sklearn-style estimators for cardinality- and budget-constrained feature selection in logistic regression. The package currently includes:
+
+- **L0L2Classifier**: L0-constrained L2-regularized logistic regression
+- **ResampledL0L2Classifier**: resampling-based feature selection with frequency-based aggregation to improve the selection stability
+
+## Installation
+
+To install the package, use the following command:
+
+```sh
+pip install l0l2learn
+```
+
+Please check the [MOSEK website](https://www.mosek.com/) to request and set up a license for the conic solver.
+
+## Quick Start
+
+### Feature Selection Without Resampling
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import L0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = L0L2Classifier(
+    b=3,
+    lambd=1.0
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+```
+
+### Feature Selection With Resampling
+
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import ResampledL0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = ResampledL0L2Classifier(
+    b=3,
+    param_grid={"lambd": [1.0]},
+    n_resamples=3
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+print("VIFs:         ", clf.variable_inclusion_frequencies_[clf.support_])
+print("MSFs:         ", clf.model_selection_frequencies_)
+```
+
+## Hyperparameters
+
+### Feature Costs
+
+Feature-specific costs can be supplied through `c`:
+
+```python
+clf = L0L2Classifier(c=[1, 2, 5])
+```
+
+The optimization then accounts for some variables to be more expensive than others.
+
+### Feature Selection Budget
+
+The feature selection budget is controlled through `b`:
+
+```python
+clf = L0L2Classifier(b=5)
+```
+
+When all feature costs are equal to one (the default), `b` directly controls the maximum number of selected features.
+
+### L2 Regularization
+
+The L2 regularization strength is given by `lambd`:
+
+```python
+clf = L0L2Classifier(lambd=0.1)
+```
+
+Larger values can attenuate overfitting and increase robustness.
+
+### Number of Resamples
+
+`n_resamples` determines how many resampled models are fitted:
+
+```python
+clf = ResampledL0L2Classifier(b=5, n_resamples=99)
+```
+
+Larger values can improve frequency estimates but increase runtime.
+
+### Other Hyperparameters
+
+#### L0L2Classifier
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`time_limit`**: Maximum runtime in seconds for the optimization problem.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+#### ResampledL0L2Classifier
+
+- **`resampling`**: Controls whether and how rows, columns, both, or neither are resampled.
+
+- **`n_row_subsamples`**: Number or fraction of observations used during row subsampling.
+
+- **`n_column_subsamples`**: Number or fraction of features used during column subsampling.
+
+- **`aggregation`**: Whether model selection or variable inclusion frequencies are used for aggregation.
+
+- **`vif_threshold`**: Minimum variable inclusion frequency required when using `aggregation="VIF"`.
+
+- **`estimator`**: Alternative base estimator used instead of the default `L0L2Classifier`.
+
+- **`param_grid`**: Hyperparameter grid used for cross-validation when tuning `lambd`.
+
+- **`cv`**: Cross-validation strategy used for hyperparameter tuning.
+
+- **`scoring`**: Scoring metric used to select the best hyperparameter configuration.
+
+- **`numerical_features`**: Specifies which DataFrame columns should be treated as numerical features.
+
+- **`categorical_features`**: Specifies which DataFrame columns should be treated as categorical features.
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`mosek_time_limit`**: Maximum runtime in seconds for each individual optimization problem.
+
+- **`total_time_limit`**: Maximum runtime in seconds for the complete resampling procedure.
+
+- **`max_consecutive_failures`**: Stops resampling if too many consecutive model fits fail.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+- **`n_jobs`**: Number of parallel workers used during resampling.
+
+- **`random_state`**: Controls the randomness of resampling and cross-validation procedures.
+
+## License
+
+This project is licensed under the [MIT License](https://github.com/ml-lab-htw/l0l2learn/blob/main/LICENSE). See the `LICENSE` file for details.
+
+## Authors
+
+- Ricardo Knauer (HTW Berlin)

l0l2learn-0.1.0/l0l2learn.egg-info/PKG-INFO

@@ -0,0 +1,213 @@
+Metadata-Version: 2.4
+Name: l0l2learn
+Version: 0.1.0
+Summary: Cardinality- and budget-constrained feature selection for logistic regression using mixed-integer conic optimization
+Home-page: https://github.com/ml-lab-htw/l0l2learn
+Author: Ricardo Knauer
+Author-email: ricardo.knauer@htw-berlin.de
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: joblib
+Requires-Dist: mosek
+Requires-Dist: numpy
+Requires-Dist: pandas
+Requires-Dist: scikit-learn
+Requires-Dist: scipy
+Requires-Dist: tqdm
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: license-file
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# l0l2learn
+
+Feature selection for logistic regression using mixed-integer conic optimization. Unlike Lasso-based approaches, `l0l2learn` directly optimizes feature subsets under explicit cardinality or budget constraints.
+
+## Overview
+
+`l0l2learn` is a Python package that provides sklearn-style estimators for cardinality- and budget-constrained feature selection in logistic regression. The package currently includes:
+
+- **L0L2Classifier**: L0-constrained L2-regularized logistic regression
+- **ResampledL0L2Classifier**: resampling-based feature selection with frequency-based aggregation to improve the selection stability
+
+## Installation
+
+To install the package, use the following command:
+
+```sh
+pip install l0l2learn
+```
+
+Please check the [MOSEK website](https://www.mosek.com/) to request and set up a license for the conic solver.
+
+## Quick Start
+
+### Feature Selection Without Resampling
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import L0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = L0L2Classifier(
+    b=3,
+    lambd=1.0
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+```
+
+### Feature Selection With Resampling
+
+
+```python
+from sklearn.datasets import load_breast_cancer
+from sklearn.metrics import roc_auc_score
+from sklearn.model_selection import train_test_split
+
+from l0l2learn import ResampledL0L2Classifier
+
+
+X, y = load_breast_cancer(return_X_y=True, as_frame=True)
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, stratify=y
+)
+
+clf = ResampledL0L2Classifier(
+    b=3,
+    param_grid={"lambd": [1.0]},
+    n_resamples=3
+)
+
+clf.fit(X_train, y_train)
+y_proba = clf.predict_proba(X_test)
+
+print("ROC AUC:      ", roc_auc_score(y_test, y_proba[:, 1]))
+print("Coefficients: ", clf.coef_)
+print("Intercept:    ", clf.intercept_)
+print("Support:      ", clf.support_)
+print("VIFs:         ", clf.variable_inclusion_frequencies_[clf.support_])
+print("MSFs:         ", clf.model_selection_frequencies_)
+```
+
+## Hyperparameters
+
+### Feature Costs
+
+Feature-specific costs can be supplied through `c`:
+
+```python
+clf = L0L2Classifier(c=[1, 2, 5])
+```
+
+The optimization then accounts for some variables to be more expensive than others.
+
+### Feature Selection Budget
+
+The feature selection budget is controlled through `b`:
+
+```python
+clf = L0L2Classifier(b=5)
+```
+
+When all feature costs are equal to one (the default), `b` directly controls the maximum number of selected features.
+
+### L2 Regularization
+
+The L2 regularization strength is given by `lambd`:
+
+```python
+clf = L0L2Classifier(lambd=0.1)
+```
+
+Larger values can attenuate overfitting and increase robustness.
+
+### Number of Resamples
+
+`n_resamples` determines how many resampled models are fitted:
+
+```python
+clf = ResampledL0L2Classifier(b=5, n_resamples=99)
+```
+
+Larger values can improve frequency estimates but increase runtime.
+
+### Other Hyperparameters
+
+#### L0L2Classifier
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`time_limit`**: Maximum runtime in seconds for the optimization problem.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+#### ResampledL0L2Classifier
+
+- **`resampling`**: Controls whether and how rows, columns, both, or neither are resampled.
+
+- **`n_row_subsamples`**: Number or fraction of observations used during row subsampling.
+
+- **`n_column_subsamples`**: Number or fraction of features used during column subsampling.
+
+- **`aggregation`**: Whether model selection or variable inclusion frequencies are used for aggregation.
+
+- **`vif_threshold`**: Minimum variable inclusion frequency required when using `aggregation="VIF"`.
+
+- **`estimator`**: Alternative base estimator used instead of the default `L0L2Classifier`.
+
+- **`param_grid`**: Hyperparameter grid used for cross-validation when tuning `lambd`.
+
+- **`cv`**: Cross-validation strategy used for hyperparameter tuning.
+
+- **`scoring`**: Scoring metric used to select the best hyperparameter configuration.
+
+- **`numerical_features`**: Specifies which DataFrame columns should be treated as numerical features.
+
+- **`categorical_features`**: Specifies which DataFrame columns should be treated as categorical features.
+
+- **`fit_intercept`**: Whether an intercept term is included in the logistic regression model.
+
+- **`mosek_time_limit`**: Maximum runtime in seconds for each individual optimization problem.
+
+- **`total_time_limit`**: Maximum runtime in seconds for the complete resampling procedure.
+
+- **`max_consecutive_failures`**: Stops resampling if too many consecutive model fits fail.
+
+- **`mosek_log`**: Enables printing of MOSEK solver output.
+
+- **`n_jobs`**: Number of parallel workers used during resampling.
+
+- **`random_state`**: Controls the randomness of resampling and cross-validation procedures.
+
+## License
+
+This project is licensed under the [MIT License](https://github.com/ml-lab-htw/l0l2learn/blob/main/LICENSE). See the `LICENSE` file for details.
+
+## Authors
+
+- Ricardo Knauer (HTW Berlin)

l0l2learn-0.1.0/l0l2learn.egg-info/SOURCES.txt

@@ -0,0 +1,8 @@
+LICENSE
+README.md
+setup.py
+l0l2learn.egg-info/PKG-INFO
+l0l2learn.egg-info/SOURCES.txt
+l0l2learn.egg-info/dependency_links.txt
+l0l2learn.egg-info/requires.txt
+l0l2learn.egg-info/top_level.txt

l0l2learn-0.1.0/l0l2learn.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

l0l2learn-0.1.0/l0l2learn.egg-info/requires.txt

@@ -0,0 +1,7 @@
+joblib
+mosek
+numpy
+pandas
+scikit-learn
+scipy
+tqdm

l0l2learn-0.1.0/l0l2learn.egg-info/top_level.txt

@@ -0,0 +1 @@
+

l0l2learn-0.1.0/setup.cfg

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

l0l2learn-0.1.0/setup.py

@@ -0,0 +1,30 @@
+from pathlib import Path
+from setuptools import setup, find_packages
+
+
+setup(
+    name="l0l2learn",
+    version="0.1.0",
+    description="Cardinality- and budget-constrained feature selection for logistic regression using mixed-integer conic optimization",
+    author="Ricardo Knauer",
+    author_email="ricardo.knauer@htw-berlin.de",
+    url="https://github.com/ml-lab-htw/l0l2learn",
+    packages=find_packages(),
+    install_requires=[
+        "joblib",
+        "mosek",
+        "numpy",
+        "pandas",
+        "scikit-learn",
+        "scipy",
+        "tqdm"
+    ],
+    long_description=Path("README.md").read_text(encoding="utf-8"),
+    long_description_content_type="text/markdown",
+    classifiers=[
+        "Programming Language :: Python :: 3",
+        "License :: OSI Approved :: MIT License",
+        "Operating System :: OS Independent",
+    ],
+    python_requires=">=3.10",
+)