Presentation

The ReDICE Consortium is a research project gathering partners from industry and academia around innovative mathematical methods for the design and analysis of computer experiments.

Partners

• UNIBE (University of Bern): Academic partner and project coordinator
• EMSE (École Nationale Supérieure des Mines de Saint-Étienne): Academic partner
• INRIA (Institut National de Recherche en Informatique et Automatique): Academic and project administration partner
• IRSN (Institut de Radioprotection et de Sûreté Nucléaire): Industry partner
• EDF R&D (Électricité de France): Industry partner
• Renault: Industry partner
• CEA (Commissariat à l'Énergie Atomique): Industry partner
• IFPen (IFP Énergies Nouvelles): Industry partner
• IMT (Institut de Mathématiques de Toulouse): Scientific expertise
• UNICE (Université de Nice): Scientific expertise
• Alpestat : Software expertise

Main research topics

• Metamodel-based optimization, inversion, and related strategies
• Multi-fidelity metamodels and mixtures of metamodels
• Computer experiments involving functional data
• Special kernels and designs

Ph.D. theses

• François Bachoc: Designs and algorithms for hyperparameter estimation (Associate CEA Ph.D. student).
• Mickaël Binois: Gaussian process-assisted multiobjective optimization (Ph.D. funded by Renault and ReDICE).
• Clément Chevalier: Kriging-based strategies for inversion under uncertainty (Ph.D. funded by IRSN and ReDICE).
• Loïc Le Gratiet: Multi-fidelity kriging for uncertainty quantification (Associate CEA Ph.D. student).
• Federico Zertuche: New multi-fidelity kriging metamodels (Ph.D. funded by Université Joseph Fourier and ReDICE).

Selected research reports and publications

• DiceKriging, DiceOptim: Two R packages for the analysis of computer experiments by kriging-based metamodelling and optimization.
  Status: In Journal of Statistical Software, 51 (1), 2012.
  Olivier Roustant, David Ginsbourger, Yves Deville.
• Estimating and quantifying uncertainties on level sets using the Vorob'ev expectation and deviation with Gaussian process models.
  Status: To appear in mODa 10 Advances in Model-Oriented Design and Analysis, Physica-Verlag HD, 2013.
  Clément Chevalier, David Ginsbourger, Julien Bect, and Ilya Molchanov.
• Fast kriging-based stepwise uncertainty reduction with application to the identification of an excursion set.
  Status: Accepted with minor revisions to Technometrics.
  Clément Chevalier, Julien Bect, David Ginsbourger, Victor Picheny, Yann Richet and Emmanuel Vazquez.
• The KrigInv package: An efficient and user-friendly R implementation of Kriging-based inversion algorithms.
  Status: To appear in Computational Statistics and Data Analysis (author version already available online here).
  Clément Chevalier, Victor Picheny, and David Ginsbourger.
• Cross Validation and Maximum Likelihood estimations of hyper-parameters of Gaussian processes with model misspecification.
  Status: To appear in Computational Statistics and Data Analysis.
  François Bachoc.
• Asymptotic analysis of the role of spatial sampling for hyper-parameter estimation of Gaussian processes.
  Status: Submitted.
  François Bachoc.
• Fast computation of the multipoint Expected Improvement with applications in batch selection.
  Status: To appear in the proceedings of the LION7 conference, Lecture Notes in Computer Science, 2013.
  Clément Chevalier and David Ginsbourger.
• Using the Efficient Global Optimization Algorithm to assist Nuclear Criticality Safety Assessment.
  Status: To appear in Nuclear Science and Engineering.
  Yann Richet, Gregory Caplin, Jérome Crevel, David Ginsbourger, and Victor Picheny.

Publicly available software (R packages)

• KrigInv: Kriging-based Inversion for Deterministic and Noisy Computer Experiments
• MuFiCokriging: Multi-Fidelity Cokriging models
• DiceKriging: Kriging methods for computer experiments (Continuation of a package initiated within the DICE Consortium)
• DiceOptim : Kriging-based optimization for computer experiments (Continuation of a package initiated within the DICE Consortium)