glmpynet 0.4.0

A scikit-learn compatible wrapper for glmnet-based logistic regression.

glmpynet

glmnet-based Logistic Regression for Scikit-Learn

glmpynet is a high-performance Python wrapper for the glmnet library, providing a scikit-learn compatible estimator for penalized logistic regression.

This project aims to bridge the gap between the raw computational speed of the original Fortran/C++ glmnet code and the ease-of-use of the Python data science ecosystem. It provides a single, focused class that acts as a drop-in replacement for sklearn.linear_model.LogisticRegression for users who need the power of elastic-net regularization for binary classification.

Key Features

• High Performance: Leverages the highly optimized, battle-tested glmnet Fortran backend for fitting models, making it suitable for large datasets.

• Scikit-learn Compatible: Implements the standard fit, predict, and predict_proba API, allowing it to be seamlessly integrated into sklearn pipelines, GridSearchCV, and other tools.

• Full Regularization Suite: Supports L1 (Lasso), L2 (Ridge), and Elastic-Net regularization for robust feature selection and prevention of overfitting.

Installation

Once released, you will be able to install glmpynet via pip:

pip install glmpynet

Quick Start

Using glmpynet is designed to be as simple as using any other scikit-learn estimator.

import numpy as np
from glmpynet import LogisticNet
from sklearn.model\_selection import train\_test\_split
from sklearn.metrics import accuracy\_score

1. Generate synthetic data

X, y = np.random.rand(100, 10), np.random.randint(0, 2, 100)
X\_train, X\_test, y\_train, y\_test = train\_test\_split(X, y)

2. Instantiate and fit the model

# alpha=1 -\> Lasso, alpha=0 -\> Ridge, 0 \< alpha \< 1 -\> Elastic-Net

model = LogisticNet(alpha=0.5)
model.fit(X\_train, y\_train)

3. Make predictions

y\_pred = model.predict(X\_test)

4. Evaluate the model

accuracy = accuracy\_score(y\_test, y\_pred)
print(f"Model Accuracy: {accuracy:.2f}")

Project Status

This project is currently in the planning and development phase. The goal is to provide a simple, robust, and well-tested wrapper for the core binary classification functionality of glmnet.

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md file for guidelines on how to report bugs, suggest features, or submit pull requests.

License

This project is distributed under the MIT License.