Third party projects and code snippets

This page references projects and code snippets (gists) which are compatible with the scikit-learn API conventions.

Projects

Different models and algorithms

• lightning Large-scale linear classification and regression in Python/Cython
• libOPF Optimal path forest classifier
• pyIPCA Incremental Principal Component Analysis
• sklearn_pandas bridge for scikit-learn pipelines and pandas data frame with dedicated transformers.
• py-earth Multivariate adaptive regression splines
• [HMMLearn] (https://github.com/hmmlearn/hmmlearn) Hidden Markov Models
• sklearn-compiledtrees generate a C++ implementation of the predict function for decision trees (and ensembles) trained by sklearn. Useful for latency-sensitive production environments.
• glm-sklearn scikit-learn compatible wrapper around the GLM module in statsmodel.
• Fast svmlight / libsvm file loader
• pyensemble An implementation of Caruana et al's Ensemble Selection algorithm in Python, based on scikit-learn
• seqlearn: sequence classification library (HMMs, structured perceptron)
• lda: fast implementation of Latent Dirichlet Allocation in Cython (github)
• random-output-trees Multi-output random forest on randomised output space
• nolearn scikit-learn compatible wrappers for neural net libraries, and other utilities.
• sklearn-theano Scikit-learn compatible tools using theano
• Sparse Filtering Unsupervised feature learning based on sparse-filtering
• Kernel Regression Implementation of Nadaraya-Watson kernel regression with automatic bandwidth selection
• Extreme Learning Machines Implementation of ELM (random layer + ridge) with a scikit-learn compatible interface.
• gplearn Genetic Programming for symbolic regression tasks.
• auto-sklearn Drop-in replacement for a scikit-learn estimator that performs automatic model and parameter selection
• fastFM Fast factorization machine implementation compatible with scikit-learn
• pyFM Another implementation of FMs in Python
• kmodes k-modes clustering algorithm for categorical data, and several of its variations
• sklearn-deap Use evolutionary algorithms instead of gridsearch in scikit-learn.
• gp-extras Additional kernels that can be used in scikit-learn's GaussianProcessRegressor

Application-specific projects

• astroML data mining for astronomy
• [nilearn] (http://nilearn.github.io/) NeuroImaging with scikit-learn

Gists

• Multi-Layer-Perceptron neural network classifier trained by SGD
• Non-Negative Garotte
• Kernel SGD
• Fuzzy K-means and K-medians
• Kernel k-means
• Non-negative Least-Squares
• Non-negative Matrix Factorization for I-divergence
• K-means + RBF transformation, inspired by The secret of the big guys
• Multiclass SVMs
• Coordinate descent solver for NMF (designed for sparse data without missing values)

Other

Code snippets that do not follow the fit / predict / transform API.

• Adaptive Lasso (no class provided)
• Generating data with non-parametric Gaussian mixture models. Useful if you need "random" data that should have non-trivial structure.
• scikit-protopy scikit-learn compatible prototype selection and generation algorithms.