Adrien Todeschini

French National Centre for Scientific Research | CNRS · Mathematical Institute of Bordeaux

Abstract: We propose a novel statistical model for sparse networks with overlapping community structure. The model is based on representing the graph as an exchangeable point process, and naturally generalizes existing probabilistic models with overlapping block-structure to the sparse regime. Our construction builds on vectors of completely random...

Nous proposons deux nouvelles approches pour les systèmes de recommandation etles réseaux. Dans la première partie, nous donnons d’abord un aperçu sur les systèmes de recommandationavant de nous concentrer sur les approches de rang faible pour la complétionde matrice. En nous appuyant sur une approche probabiliste, nous proposons de nouvellesfoncti...

We propose two novel approaches for recommender systems and networks. In the first part, we first give an overview of recommender systems and concentrate on the low-rank approaches for matrix completion. Building on a probabilistic approach, we propose novel penalty functions on the singular values of the low-rank matrix. By exploiting a mixture mo...

We propose a novel statistical model for sparse networks with overlapping community structure. The model is based on representing the graph as an exchangeable point process, and naturally generalizes existing probabilistic models with overlapping block-structure to the sparse regime. Our construction builds on vectors of completely random measures,...

Nous proposons un modèle bayésien non paramétrique (BNP) à faible rang pour les graphes bipartis. Récemment, Caron (2012) a proposé un modèle BNP où chaque élément possède son propre paramètre de sociabilité permettant de capturer le compor-tement en loi de puissance observé dans les graphes bipartis réels. Ce modèle peut être considéré comme une f...

Biips is a software platform for automatic Bayesian inference with interacting particle systems. Biips allows users to define their statistical model in the probabilistic programming BUGS language, as well as to add custom functions or samplers within this language. Then it runs sequential Monte Carlo based algorithms (particle filters, particle in...

Nous proposons une nouvelle classe d'algorithmes pour la complétion de matrice de rang faible. Notre approche s'appuie sur de nouvelles fonctions de pénalité sur les valeurs singulières de la matrice de rang faible. En exploitant une représentation basée sur un modèle de mélange de cette pénalité, nous montrons qu'un ensemble de variables latentes...

We consider the problem of local radioelectric property estimation from global electromagnetic scattering measurements. This challenging ill-posed high dimensional inverse problem can be explored by intensive computations of a parallel Maxwell solver on a petaflopic supercomputer. Then, it is shown how Bayesian inference can be perfomed with a Part...

We propose a novel class of algorithms for low rank matrix completion. Our approach builds on novel penalty functions on the singular values of the low rank matrix. By exploiting a mixture model representation of this penalty, we show that a suitably chosen set of latent variables enables to derive an Expectation-Maximization algorithm to obtain a...