wrap-glmnet
wrap_glmnet is a small Python wrapper around python-glmnet's
LogitNet classifier. It keeps the glmnet elastic-net logistic regression
solver, but exposes a more sklearn-like estimator and adds control over
glmnet's internal cross-validation behavior.
Why it exists
python-glmnet is useful for regularized logistic regression, but its
classifier API has a few rough edges for sklearn workflows. This package
provides GlmnetLogitNetWrapper, which is intended to be used in place of
glmnet.LogitNet when you need:
- sklearn-compatible cloning and fitted attributes such as
n_features_in_andfeature_names_in_ - explicit selection of whether predictions use
lambda_1seor the best-performing cross-validation lambda - sklearn-style class weights
- custom internal cross-validation splitters, including group-aware splitters
- multiclass ROC-AUC and deviance/log-loss scorers for glmnet's internal CV
- held-out CV predicted probabilities and per-fold CV scores
- predictable
decision_functionandpredict_probaoutput shapes for binary, multiclass, single-lambda, and multi-lambda predictions
How it works
Importing wrap_glmnet patches selected python-glmnet internals so glmnet's
lambda-path scoring can use an optional sklearn-style internal_cv splitter.
The patched scoring path also forwards groups to scorers that accept it,
records _cv_scores_, and, by default, stores held-out predicted probabilities
in cv_pred_probs_ with shape (n_samples, n_classes, n_lambdas).
GlmnetLogitNetWrapper delegates the actual model fit to an inner
glmnet.LogitNet instance. Constructor keyword arguments not handled by the
wrapper are passed through to LogitNet.
By default, predictions, coef_, intercept_, cv_mean_score_final_, and
cv_standard_error_final_ use lambda_1se (lambda_best_ in
python-glmnet). Set use_lambda_1se=False to use the lambda with the best
mean CV score (lambda_max_), or call switch_lambda(...) after fitting to
get a copied fitted model with the other lambda choice.
Installation
The package requires Python 3.10 or newer.
pip install wrap_glmnet
For local development:
pip install -r requirements_dev.txt
pip install -e .
Usage
from sklearn.model_selection import StratifiedGroupKFold
from wrap_glmnet import GlmnetLogitNetWrapper
clf = GlmnetLogitNetWrapper(
alpha=1.0,
n_lambda=100,
internal_cv=StratifiedGroupKFold(n_splits=3),
scoring=GlmnetLogitNetWrapper.rocauc_scorer,
require_cv_group_labels=True,
)
clf.fit(X_train, y_train, groups=groups)
labels = clf.predict(X_test)
probabilities = clf.predict_proba(X_test)
# Compare the default lambda_1se model to the best-CV-score lambda.
best_cv_lambda_clf = clf.switch_lambda(use_lambda_1se=False)
best_cv_probabilities = best_cv_lambda_clf.predict_proba(X_test)
If internal_cv is not supplied, the wrapper uses glmnet's normal internal CV
setup with n_splits (default: 3). If require_cv_group_labels=True, calling
fit(...) without groups raises an error.
class_weight accepts sklearn-style values such as a class-to-weight mapping or
"balanced". When both class_weight and sample_weight are supplied, the
wrapper multiplies them and normalizes the resulting sample weights before
fitting.
Important behavior and limitations
- This package is focused on
glmnet.LogitNetclassification, not every glmnet model type. - It patches
python-glmnetfunctions at import time, so it relies onpython-glmnetinternals remaining compatible. store_cv_predicted_probabilities=Trueis the default and can use substantial memory for large datasets or long lambda paths. Disable it if you do not needcv_pred_probs_.- Multiclass probabilities are computed from the wrapper's decision scores with softmax normalization; binary probabilities use sigmoid-style normalization.
Development
make test
make lint
make docs
Tests cover sklearn cloning, scorer behavior, group-aware CV, lambda switching, stored CV outputs, plotting, and output shapes.
Changelog
0.0.1
- First release on PyPI.
Metadata
Release files for wrap-glmnet 0.1.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| wrap_glmnet-0.1.3.tar.gz | 26.0 kB | Without provenance | Details for wrap_glmnet-0.1.3.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| wrap_glmnet-0.1.3-py2.py3-none-any.whl | Python 3, Python 2 | none | any | Details for wrap_glmnet-0.1.3-py2.py3-none-any.whl |
Total release size: 40.0 kB
Release files / wrap_glmnet-0.1.3.tar.gz
| Download URL | wrap_glmnet-0.1.3.tar.gz |
|---|---|
| Size | 26.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
7933bed3b8b278dc6b51af6e25c7354cc16107971bb84676a4c743542d523561
|
|
BLAKE2b-256 checksum How to use checksums |
2b9309084eb493afd82a3f0f3c86d4eddd0c0505bc6911977eb77b2150a0c096
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.25
|
Release files / wrap_glmnet-0.1.3-py2.py3-none-any.whl
| Download URL | wrap_glmnet-0.1.3-py2.py3-none-any.whl |
|---|---|
| Size | 14.0 kB |
| Tags | Python 2 Python 3 |
|
SHA-256 checksum How to use checksums |
af345c81b6c83fbad2c837221fe6c41263df142ebb39ed3e8f05af7d72c7f95c
|
|
BLAKE2b-256 checksum How to use checksums |
8933839c1f40d87940b38df88d88d117310ccdb07f054873bbfba319aaddcacc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.25
|
Security for wrap-glmnet 0.1.3
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.