A Chaining Algorithm for Online Nonparametric Regression

We consider the problem of online nonparametric regression with arbitrary deterministic sequences. Using ideas from the chaining technique, we design an algorithm that achieves a Dudley-type regret bound similar to the one obtained in a non-constructive fashion by Rakhlin and Sridharan (2014). Our regret bound is expressed in terms of the metric entropy in the sup norm, which yields optimal guarantees when the metric and sequential entropies are of the same order of magnitude. In particular our algorithm is the first one that achieves optimal rates for online regression over Hölder balls.

Workshop "Sequential Learning and Applications" à Toulouse les 9-10 novembre prochains

Le Labex CIMI organise actuellement un trimestre thématique dédié à l'apprentissage (machine learning) à Toulouse. Dans ce cadre, le projet SPADRO est associé pour l'organisation d'un workshop les 9 et 10 novembre prochains, dans lequel les bandits (et l'analyse convexe) joueront un rôle central : voir la page de l'événement.

Emilien Joly rejoint SPADRO!

L'équipe SPADRO souhaite chaleureusement la bienvenue à Emilien Joly!

Emilien est un spécialiste de l'estimation robuste, de la concentration, des inégalités isopérimétriques et de l'hypercontractivité. Il a préparé son doctorat de statistique sous la direction jointe de Gabor Lugosi et Gilles Stoltz. Cette année universitaire, Emilien se consacrera au projet post-doctoral d'étude des bandits multiples depuis la perspective de l'inférence semiparamétrique.

Empirial ϕ∗-discrepancies and quasi-empirical likelihood: exponential bounds

We review some recent extensions of the so-called generalized empirical likelihood method, when the Kullback distance is replaced by some general convex divergence. We propose to use, instead of empirical likelihood, some regularized form or quasi-empirical likelihood method, corresponding to a convex combination of Kullback and chi2 discrepancies. We show that for some adequate choice of the weight in this combination, the corresponding quasi-empirical likelihood is Bartlett-correctable.

Drawing valid targeted inference when covariate-adjusted response-adaptive RCT meets data-adaptive loss-based estimation, with an application to the LASSO

Adaptive clinical trial design methods have garnered growing attention in the recent years, in large part due to their greater flexibility over their traditional counterparts. One such design is the so-called covariate-adjusted, response-adaptive (CARA) randomized controlled trial (RCT).

On the Complexity of Best Arm Identification in Multi-Armed Bandit Models

The stochastic multi-armed bandit model is a simple abstraction that has proven useful in many different contexts in statistics and machine learning. Whereas the achievable limit in terms of regret minimization is now well known, our aim is to contribute to a better understanding of the performance in terms of identifying the m best arms. We introduce generic notions of complexity for the two dominant frameworks considered in the literature: fixed-budget and fixed-confidence settings.

Exposé d'Antoine Chambaz au Séminaire Parisien de Statistique, lundi 13 avril, 15h, à l'IHP.

Antoine Chambaz présentera un exposé intitulé
Intervalles de confiance pour les bandits contextuels
au Séminaire Parisien de Statistique, lundi 13 avril, 15h, à l'IHP.

Voici le résumé de son intervention:

Emilie Kaufmann lauréate du Prix de thèse Jacques Neveu 2014!

Félicitations à Emilie Kaufmann, lauréate du Prix de thèse Jacques Neveu 2014!



Empirical phi star-Divergence Minimizers for Hadamard Differentiable Functionals

We study some extensions of the empirical likelihood method, when the Kullback distance is replaced by some general convex divergence or phi-discrepancy. We show that this generalized empirical likelihood method is asymptotically valid for general Hadamard differentiable functionals.

La présentation de Pierre Barbillon est disponible

Voici les transparents de la présentation de Pierre Barbillon lors de la réunion SPADRO du 23 janvier 2015.

Merci beaucoup à lui de les mettre à disposition de l'équipe.


Souscrire à spadro.eu RSS