Aurélien Bellet

Open-source libraries

• Metric-learn [GitHub] [PyPI] [Documentation]
  Python toolbox implementing various metric learning algorithms, among which Large Margin Nearest Neighbor (LMNN), Neighborhood Component Analysis (NCA), Information-Theoretic Metric Learning (ITML), Relative Component Analysis (RCA) and Mahalanobis Metric for Clustering (MMC). The API is compatible with scikit-learn, allowing easy model selection and evaluation. Distributed under MIT license.

Open-source code from papers

• Distributed Frank Wolfe for Low-Rank Learning [GitHub]
  Python / PySpark implementation of the distributed optimization algorithm proposed in our paper A Distributed Frank-Wolfe Framework for Learning Low-Rank Matrices with the Trace Norm (Machine Learning 2018). The code can run on Spark clusters and allows user to define their own objective functions.
• GoSta [GitHub]
  Basic Python implementation of the gossip algorithm for computing U-statistics proposed in our paper Extending Gossip Algorithms to Distributed Estimation of U-statistics (NIPS 2015). A few other gossip averaging algorithms are also implemented. Distributed under Apache 2.0 license.
• Distributed Frank Wolfe for Sparse Learning [mloss]
  C++/MPI implementation of the distributed optimization algorithm proposed in our paper A Distributed Frank-Wolfe Algorithm for Communication-Efficient Sparse Learning (SDM 2015). In its current version, it can tackle two problems: kernel SVM with distributed training examples and LASSO regression with distributed attributes. Distributed under GNU/GPL 3 license.
• SCML [GitHub] [mloss]
  Matlab/MEX implementation of the metric learning method proposed in our paper Sparse Compositional Metric Learning (AAAI 2014). It allows scalable learning of global, multi-task and multiple local Mahalanobis metrics for multi-class data under a unified framework based on sparse combinations of rank-one basis metrics. Distributed under GNU/GPL 3 license.
• GESL [mloss]
  C library implementing the string edit similarity learning method proposed in our papers Learning Good Edit Similarities with Generalization Guarantees (ECML/PKDD 2011) and Good edit similarity learning by loss minimization (MLJ 2012). It is able to handle multi-class problems and also includes a classification library that allows to learn a sparse classifier from an edit similarity. The archive includes the datasets used in the papers. Distributed under GNU/GPL 3 license.