stochs: C++ library with Cython bindings for fast stochastic optimization algorithms for machine learning, including stochastic approximation, variance reduction and hybrid algorithms like stochastic MISO (Bietti and Mairal, 2017) for handling finite datasets with random perturbations.

Vowpal Wabbit: Multiple contributions, mostly related to Contextual Bandit algorithms (see Bietti, Agarwal and Langford, 2018).

RKHS regularization: Pytorch package with regularizers based on approximations of the RKHS norm of generic CNNs, for small datasets and adversarial robustness (see Bietti, Mialon, Chen and Mairal, 2019).

convolutional kernels: C++/Cython package for computing exact kernel evaluations for convolutional kernels (CKN and NTK).

qmf: C++ library for implicit-feedback matrix factorization algorithms. Built with Denis Yarats while at Quora.

online HMM: Python/numpy code for online EM algorithms in hidden Markov and semi-Markov models (see Bietti, Bach and Cont, 2015).

infimnist-py: Cython extension for L. Bottou’s infinite MNIST dataset generator, along with a Tensorflow queue for real-time data generation.

dpmm: Variational inference for Dirichlet process mixtures of multinomials, e.g. for text document clustering, in Python/numpy (accompanying report here).