pyglmnet

A python implementation of elastic-net regularized generalized linear models

[Documentation (stable version)] [Documentation (development version)]

Pyglmnet provides a wide range of noise models (and paired canonical link functions): 'gaussian', 'binomial', 'probit', 'gamma', 'poisson', and 'softplus'.
It supports a wide range of regularizers: ridge, lasso, elastic net, group lasso, and Tikhonov regularization.
We have implemented a cyclical coordinate descent optimizer with Newton update, active sets, update caching, and warm restarts. This optimization approach is identical to the one used in R package.
A number of Python wrappers exist for the R glmnet package (e.g. here and here) but in contrast to these, Pyglmnet is a pure python implementation. Therefore, it is easy to modify and introduce additional noise models and regularizers in the future.

Installation

Install the stable PyPI version with pip

$ pip install pyglmnet

For the bleeding edge development version:

Clone the repository.

$ pip install https://api.github.com/repos/glm-tools/pyglmnet/zipball/master

Getting Started

Here is an example on how to use the GLM estimator.

import numpy as np
import scipy.sparse as sps
from pyglmnet import GLM, simulate_glm

n_samples, n_features = 1000, 100
distr = 'poisson'

# sample a sparse model
beta0 = np.random.rand()
beta = np.random.random(n_features)
beta[beta < 0.9] = 0

# simulate data
Xtrain = np.random.normal(0.0, 1.0, [n_samples, n_features])
ytrain = simulate_glm('poisson', beta0, beta, Xtrain)
Xtest = np.random.normal(0.0, 1.0, [n_samples, n_features])
ytest = simulate_glm('poisson', beta0, beta, Xtest)

# create an instance of the GLM class
glm = GLM(distr='poisson', score_metric='deviance')

# fit the model on the training data
glm.fit(Xtrain, ytrain)

# predict using fitted model on the test data
yhat = glm.predict(Xtest)

# score the model on test data
deviance = glm.score(Xtest, ytest)

More pyglmnet examples and use cases.

Tutorial

Here is an extensive tutorial on GLMs, optimization and pseudo-code.

Here are slides from a talk at PyData Chicago 2016, corresponding tutorial notebooks and a video.

How to contribute?

We welcome pull requests. Please see our developer documentation page for more details.

Acknowledgments

Konrad Kording for funding and support
Sara Solla for masterful GLM lectures

kjsandbrink/pyglmnet