Yvann Le Fay's Projects
A java implementation of the general abelian sandpile dynamic.
Efficient SDE samplers including Gaussian-based probabilistic solvers. Written in JAX.
exercices de mathématiques sup et spé, colles et exos classiques, Lycée Henri-IV 2018-2020.
JAX implementation of the Greenkhorn and adaptive primal-dual accelerated mirror descent (APDAMD) Algorithms. Project done as part of the course on computional optimal transport by Gabriel Peyré.
Implementation of High-dimensional vector autoregression time series modeling via tensor decomposition, Di Wang, Yao Zheng, Heng Lian, Guodong Li. Written in JAX.
Project done as part of the course on Bayesian Statistics, Anna Simoni. Implementation of a block Gibbs sampler.
Project done as part of the course on Probabilistic Graphical Models by Pierre Latouche and Pierre-Alexandre Mattei. JAX Implementation of the FastICA algorithm, a Newton's descent algorithm for linear ICA, and a Flax implementation of VAE for non-linear ICA.
Neurolang enables the analysis of NeuroImaging data through probabilistic logic programming. It seamlessly allow to combine, images, databases, and ontologies within a single framework.
A NMT pipeline between english and Sparql/Datalog queries.
Tours on OT. Course by Gabriel Peyré.
Poisson Network Models
Numerical approximations for stochastic differential equations.
Implementation of the Carr-Madan formula for fast derivative pricing of European options.
Barrier methods for quadratic programs, project done as part of the course on Convex Optimization by Alexandre d'Aspremont. Written in JAX.
A generic library for linear and non-linear Gaussian smoothing problems. The code leverages JAX and implements several linearization algorithms, both in a sequential and parallel fashion, as well as efficient gradient rules for computing gradients of required quantities (such as the pseudo-loglikelihood of the system).
Projects and assignments done during the course of Laurent Oudre on time series. 1) Footstep classifier between healthy and non-healthy individuals using dynamic time warping (DTW) distance. 2) Orthogonal Matching Pursuit algorithm
Implementation of a coupled Metropolis-Hasting Algorithm in Jax. Project done as part of the Bayesian Machine Learning course by Rémi Bardenet and Julyan Arbel.
Bayesian world growth model