Open Source Software

The Laboratory for Computational and Statistical Learning strives to make the code for reproducing experiments free and open to everyone. Below are links to the code repositories for some of our projects. More can be found on our GitHub page.

• CRiSP for Misspecified Robot Model

  by Gian Maria Marconi, Raffaello Camoriano, Lorenzo Rosasco, and Carlo Ciliberto

  Code accompanying the paper "Structured Prediction for CRiSP Inverse Kinematics Learning with Misspecified Robot Models"

• Iterreg: Iterative Regularization For Linear Models

  by C. Molinari, M. Massias, L. Rosasco and S. Villa

  Iterreg is a scikit-learn compatible python package to perform iterative regularization of linear models. It implements the algorithm of "Iterative regularization for convex regularizers" by C. Molinari, M. Massias, L. Rosasco and S. Villa, AISTATS 2021.

• Falkon 2.0: Scaling Nyström KRR to Billions of Points

  by Giacomo Meanti, Luigi Carratino, Lorenzo Rosasco, Alessandro Rudi

  Python package implementing the Falkon algorithm for large-scale, approximate kernel ridge regression. The implementation is based on PyTorch and runs on CPU and GPU.

• Batch-BKB: Near-linear time, no regret Bayesian optimization

  by Calandriello Daniele, Luigi Carratino, Alessandro Lazaric, Michal Valko and Lorenzo Rosasco,

  Python code implementing Batch-BKB: the first Bayesian optimization (a.k.a. Gaussian process or bandit optimization) algorithm that is both provably no-regret and guaranteed to run in near-linear time time.

• DPP-VFX: Very Fast and eXact DPP sampler

  by Michał Dereziński, Calandriello Daniele, and Michal Valko

  Material related to the NeurIPS 2019 paper "Exact sampling of determinantal point processes with sublinear time preprocessing" by Michał Dereziński, Calandriello Daniele, and Michal Valko.

• BLESS: Bottom-up leverage score sampling

  by Alessandro Rudi, Daniele Calandriello, Luigi Carratino, Lorenzo Rosasco

  Python code implementing the ridge leverage score sampling algorithm BLESS presented in: On Fast Leverage Score Sampling and Optimal Learning (NIPS 2018). The implementation can exploit both GPU and CPU resources.

• FALKON: An Optimal Large Scale Kernel Method

  by Alessandro Rudi, Luigi Carratino, Lorenzo Rosasco

  This Matlab code provides an implementation of the FALKON algorithm presented in 'FALKON: An Optimal Large Scale Kernel Method' NIPS 2017, together with scripts to reproduce the experiments in the paper.

• iCubWorld - Dataset for Visual Recognition in Robotics

  by Giulia Pasquale, Carlo Ciliberto, Sean R. Fanello, Lorenzo Natale, Francesca Odone, Giorgio Metta and Lorenzo Rosasco

  The iCub World Dataset is an ongoing project in collaboration with the iCub Facility and the SLIP-GURU group at the University of Genoa. The goal of the project is to build a growing dataset for visual recognition in robotics. The dataset was acquired during Human-Robot Interaction sessions, where a human teacher showed different everyday objects to the iCub robot.

• MultiplePassesSGM: Multiple passes stochastic gradient method

  by Junhong Lin, Raffaello Camoriano and Lorenzo Rosasco

  This is the experimental code used in 'Generalization Properties and Implicit Regularization for Multiple Passes SGM', appearing in ICML 2016. http://jmlr.org/proceedings/papers/v48/lina16.pdf.

• NystromCoRe: Nyström Computational Regularization

  by Raffaello Camoriano, Alessandro Rudi and Lorenzo Rosasco

  This Matlab package provides an implementation of the Nyström Computational Regularization algorithm presented in the following work: Alessandro Rudi, Raffaello Camoriano, Lorenzo Rosasco, 'Less is More: Nyström Computational Regularization', 16 Jul 2015, http://arxiv.org/abs/1507.04717.

• NYTRO: NYström iTerative RegularizatiOn

  by Tomas Angles*, Raffaello Camoriano*, Alessandro Rudi and Lorenzo Rosasco

  This Matlab package provides an implementation of the NYTRO algorithm presented in the following work: Tomas Angles, Raffaello Camoriano, Alessandro Rudi, Lorenzo Rosasco, 'NYTRO: When Subsampling Meets Early Stopping', 19 Oct 2015, http://arxiv.org/abs/1510.05684. In NYTRO, we combine early stopping and subsampling ideas, proposing a form of randomized iterative regularization based on early stopping and subsampling. In this way, we overcome the memory bottle neck of exact Early Stopping algorithms such as the kernelized Landweber iteration. Moreover, NYTRO can also be faster than other subsampled algorithms, such as Nyström Kernel Regularized Least Squares (NKRLS), especially when a stopping rule is used.

• GURLS - Grand Unified Regularized Least Squares

  by Andrea Tacchetti, Pavan K Mallapragada, Matteo Santoro and Lorenzo Rosasco

  GURLS is a least squares, modular, easy-to-extend software library for efficient supervised learning. GURLS is targeted to machine learning practitioners, as well as non-specialists. It offers a number state-of-the-art training strategies for medium and large-scale learning, and routines for efficient model selection. The library is particularly well suited for multi-output problems (multi-category/multi-label).

• Scikit-Tensor

  by Maximilian Nickel

  Scikit-tensor is a Python module for multilinear algebra and tensor factorizations.