Pascal Massart
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Pascal Massart has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Stéphane Boucheron
UMR 7599 LPMA CNRS Université Paris-Diderot 5 rue Thomas Mann 75013 Paris France
2 shared publications - Benjamin Goehry
Laboratoire de Mathématiques d'Orsay Université Paris-Saclay France
1 shared publication - Claire Lacour
Univ Gustave Eiffel, Univ Paris Est Creteil, CNRS, LAMA UMR8050 F-77447 Marne-la-Vallée, France
1 shared publication - Gábor Lugosi
Department of Economics and Business Pompeu Fabra University Barcelona Spain
1 shared publication - Hui Yan
EDF Lab, 7 bd Gaspard Monge, 91120 Palaiseau, France
1 shared publication - Jean-Michel Poggi
LMO, Université Paris-Saclay, Fac. Sciences Orsay, Bat. 307, 91400 Orsay, France
1 shared publication
Latest Publications
- Open Access
Random Forests for Time Series
Random forests were introduced in 2001 by Breiman and have since become a popular learning algorithm, for both regression and classification. However, when dealing with time series, random forests do not integrate the time-dependent structure, implicitly supposing that the observations are independent. We propose some variants of the random forests designed for time series. The idea is to replace the standard bootstrap with a dependent bootstrap (i.e block bootstrap) to subsample time series during the tree construction phase to take time dependence into account. We then present two numerical experiments on electricity load forecasting. The first one, at a disaggregated level, is based on an application to load forecasting of a building and illustrate how the variants may perform. The second one is at a more aggregated level (French national forecasting) but focusing on atypical periods. For both, we explore a heuristic for the choice of the block size, the new parameter.
Open AccessRandom Forests for Time Series
Random forests were introduced in 2001 by Breiman and have since become a popular learning algorithm, for both regression and classification. However, when dealing with time series, random forests do not integrate the time-dependent structure, implicitly supposing that the observations are independent. We propose some variants of the random forests designed for time series. The idea is to replace the standard bootstrap with a dependent bootstrap (i.e block bootstrap) to subsample time series during the tree construction phase to take time dependence into account. We then present two numerical experiments on electricity load forecasting. The first one, at a disaggregated level, is based on an application to load forecasting of a building and illustrate how the variants may perform. The second one is at a more aggregated level (French national forecasting) but focusing on atypical periods. For both, we explore a heuristic for the choice of the block size, the new parameter.
- 1 December 2016
- Published by Elsevier in Stochastic Processes and their Applications
- Vol. 126 (12) , 3774-3789
Minimal penalty for Goldenshluger–Lepski method
- 1 December 2016
- Published by Elsevier in Stochastic Processes and their Applications
- Vol. 126 (12) , 3774-3789
Minimal penalty for Goldenshluger–Lepski method
- Open Access
A new V-fold type procedure based on robust tests
We define a general V-fold cross-validation type method based on robust tests, which is an extension of the hold-out defined by Birg{\'e} [7, Section 9]. We give some theoretical results showing that, under some weak assumptions on the considered statistical procedures, our selected estimator satisfies an oracle type inequality. We also introduce a fast algorithm that implements our method. Moreover we show in our simulations that this V-fold performs generally well for estimating a density for different sample sizes, and can handle well-known problems, such as binwidth selection for histograms or bandwidth selection for kernels. We finally provide a comparison with other classical V-fold methods and study empirically the influence of the value of V on the risk.
Open AccessA new V-fold type procedure based on robust tests
We define a general V-fold cross-validation type method based on robust tests, which is an extension of the hold-out defined by Birg{\'e} [7, Section 9]. We give some theoretical results showing that, under some weak assumptions on the considered statistical procedures, our selected estimator satisfies an oracle type inequality. We also introduce a fast algorithm that implements our method. Moreover we show in our simulations that this V-fold performs generally well for estimating a density for different sample sizes, and can handle well-known problems, such as binwidth selection for histograms or bandwidth selection for kernels. We finally provide a comparison with other classical V-fold methods and study empirically the influence of the value of V on the risk.
- 11 March 2010
- Published by Springer Nature in Probability Theory and Related Fields
- Vol. 150 (3-4) , 405-433
A high-dimensional Wilks phenomenon
- 11 March 2010
- Published by Springer Nature in Probability Theory and Related Fields
- Vol. 150 (3-4) , 405-433
A high-dimensional Wilks phenomenon
- Open Access
Data-driven calibration of penalties for least-squares regression
Penalization procedures often suffer from their dependence on multiplying factors, whose optimal values are either unknown or hard to estimate from the data. We propose a completely data-driven calibration algorithm for this parameter in the least-squares regression framework, without assuming a particular shape for the penalty. Our algorithm relies on the concept of minimal penalty, recently introduced by Birge and Massart (2007) in the context of penalized least squares for Gaussian homoscedastic regression. On the positive side, the minimal penalty can be evaluated from the data themselves, leading to a data-driven estimation of an optimal penalty which can be used in practice; on the negative side, their approach heavily relies on the homoscedastic Gaussian nature of their stochastic framework. The purpose of this paper is twofold: stating a more general heuristics for designing a data-driven penalty (the slope heuristics) and proving that it works for penalized least-squares regression with a random design, even for heteroscedastic non-Gaussian data. For technical reasons, some exact mathematical results will be proved only for regressogram bin-width selection. This is at least a first step towards further results, since the approach and the method that we use are indeed general.
Open AccessData-driven calibration of penalties for least-squares regression
Penalization procedures often suffer from their dependence on multiplying factors, whose optimal values are either unknown or hard to estimate from the data. We propose a completely data-driven calibration algorithm for this parameter in the least-squares regression framework, without assuming a particular shape for the penalty. Our algorithm relies on the concept of minimal penalty, recently introduced by Birge and Massart (2007) in the context of penalized least squares for Gaussian homoscedastic regression. On the positive side, the minimal penalty can be evaluated from the data themselves, leading to a data-driven estimation of an optimal penalty which can be used in practice; on the negative side, their approach heavily relies on the homoscedastic Gaussian nature of their stochastic framework. The purpose of this paper is twofold: stating a more general heuristics for designing a data-driven penalty (the slope heuristics) and proving that it works for penalized least-squares regression with a random design, even for heteroscedastic non-Gaussian data. For technical reasons, some exact mathematical results will be proved only for regressogram bin-width selection. This is at least a first step towards further results, since the approach and the method that we use are indeed general.
- Top Cited PapersOpen Access
- 1 July 2003
- Published by Institute of Mathematical Statistics in The Annals of Probability
- Vol. 31 (3) , 1583-1614
Concentration inequalities using the entropy method
We investigate a new methodology, worked out by Ledoux and Massart, to prove concentration-of-measure inequalities. The method is based on certain modified logarithmic Sobolev inequalities. We provide some very simple and general ready-to-use inequalities. One of these inequalities may be considered as an exponential version of the Efron--Stein inequality. The main purpose of this paper is to point out the simplicity and the generality of the approach. We show how the new method can recover many of Talagrand's revolutionary inequalities and provide new applications in a variety of problems including Rademacher averages, Rademacher chaos, the number of certain small subgraphs in a random graph, and the minimum of the empirical risk in some statistical estimation problems.
- 1 July 2003
- Published by Institute of Mathematical Statistics in The Annals of Probability
- Vol. 31 (3) , 1583-1614
Top Cited PapersOpen AccessConcentration inequalities using the entropy method
We investigate a new methodology, worked out by Ledoux and Massart, to prove concentration-of-measure inequalities. The method is based on certain modified logarithmic Sobolev inequalities. We provide some very simple and general ready-to-use inequalities. One of these inequalities may be considered as an exponential version of the Efron--Stein inequality. The main purpose of this paper is to point out the simplicity and the generality of the approach. We show how the new method can recover many of Talagrand's revolutionary inequalities and provide new applications in a variety of problems including Rademacher averages, Rademacher chaos, the number of certain small subgraphs in a random graph, and the minimum of the empirical risk in some statistical estimation problems.