Upcoming: ICML 2019 tutorial!

A Primer on PAC-Bayesian Learning

NIPS 2017 workshop

(Almost) 50 Shades of Bayesian Learning: PAC-Bayesian trends and insights

Short Bio

Since December 2018, I am a Principal Research Scientist at University College London (UCL), in the Computer Science department.

Since November 2014, I am a (tenured) researcher at Inria, member of the MODAL project-team (MOdels for Data Analysis and Learning) of the Inria Lille - Nord Europe research centre, France. I am also affiliated with the Laboratoire Paul Painlevé (UMR CNRS 8524), which is the mathematics department of the University of Lille.

I obtained a Ph.D. in mathematics in 2013 from UPMC (Université Pierre & Marie Curie, France) under the supervision of Gérard Biau and Éric Moulines. Prior to that, I was a research assistant at DTU Compute (Denmark) supervised by Gilles Guillot.

News

  • Summer 2019

    This website will soon be (largely) updated. Stay tuned! Old news have been removed.

Research

My main line of research is in statistical machine learning. I am primarily interested in the design, analysis and implementation of statistical learning methods for high dimensional problems. My interests include (but are not limited to): PAC-Bayesian theory, sparsity and high-dimensional statistics, optimisation theory, statistical learning theory, non-negative matrix factorisation, aggregation of estimators and classifiers, MCMC algorithms, (un)supervised learning, online clustering, concentration inequalities...

  • 18 Research Articles
    (9 published)
  • 6 Softwares
    (2 as sole author)
  • 70 Talks since 2010
  • 19 Students
    (3 Ph.D.)
  • 4 Collaborations with companies

Publications

arXiv • dblp • HAL • Iris • Google Scholar • ResearchGate

Preprints

  1. Z. Mhammedi, P. Grünwald and B. Guedj. PAC-Bayes Un-Expected Bernstein Inequality.
    [ pdf ] • [ link ]
  2. G. Letarte, P. Germain, B. Guedj and F. Laviolette. Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks.
    [ pdf ] • [ link ]
  3. J. M. Zhang, E. T. Barr, B. Guedj, M. Harman and J. Shawe-Taylor. Perturbed Model Validation: A New Framework to Validate Model Relevance.
    [ pdf ] • [ link ]
  4. B. Guedj and J. Rengot. Non-linear aggregation of filters to improve image denoising.
    [ pdf ] • [ link ]
  5. Stéphane Chretien and B. Guedj. Revisiting clustering as matrix factorisation on the Stiefel manifold.
    [ pdf ] • [ link ]
  6. B. Guedj. A Primer on PAC-Bayesian learning.
    [ pdf ] • [ link ]
  7. J. Klein, M. Albardan, B. Guedj and O. Colot. Decentralized learning with budgeted network load using Gaussian copulas and classifier ensembles.
    [ pdf ] • [ link ] • [ software ]
  8. B. Guedj and L. Li. Sequential Learning of Principal Curves: Summarizing Data Streams on the Fly.
    [ pdf ] • [ link ]
  9. A. Celisse and B. Guedj. Stability revisited: new generalisation bounds for the Leave-one-Out.
    [ pdf ] • [ link ]

International peer-reviewed journals

  1. L. Li, B. Guedj and S. Loustau (2018). A Quasi-Bayesian Perspective to Online Clustering.
    Electronic Journal of Statistics, vol. 12 (2), 3071--3113.
    [ pdf ] • [ link ] • [ journal website ] • [ software ]
  2. B. Guedj and B. Srinivasa Desikan (2018). Pycobra: A Python Toolbox for Ensemble Learning and Visualisation.
    Journal of Machine Learning Research, vol. 18 (190), 1--5.
    [ pdf ] • [ link ] • [ journal website ] • [ software ]
  3. P. Alquier and B. Guedj (2018). Simpler PAC-Bayesian Bounds for Hostile Data.
    Machine Learning, vol. 107 (5), 887--902.
    [ pdf ] • [ link ] • [ journal website ]
  4. B. Guedj and S. Robbiano (2018). PAC-Bayesian High Dimensional Bipartite Ranking.
    Journal of Statistical Planning and Inference, vol. 196, 70--86.
    [ pdf ] • [ link ] • [ journal website ]
  5. P. Alquier and B. Guedj (2017). An Oracle Inequality for Quasi-Bayesian Non-Negative Matrix Factorization.
    Mathematical Methods of Statistics, vol. 26(1), 55-67.
    Errata: there is a small mistake in the proofs of the published version. The latest arXiv version corrects this mistake.
    [ pdf ] • [ link ] • [ journal website ] • [ software ]
  6. G. Biau, A. Fischer, B. Guedj and J. D. Malley (2016). COBRA: A Combined Regression Strategy.
    Journal of Multivariate Analysis, vol. 146, 18-28.
    [ pdf ] • [ supplementary material ] • [ journal website ] • [ software: R package ] • [ software: Python pycobra library ]
  7. N. Chopin, S. Gadat, B. Guedj, A. Guyader and E. Vernet (2015). On some recent advances on high dimensional Bayesian statistics.
    ESAIM: Proceedings & Surveys, vol. 51, 293-319.
    [ pdf ] • [ journal website ]
  8. B. Guedj and P. Alquier (2013). PAC-Bayesian Estimation and Prediction in Sparse Additive Models.
    Electronic Journal of Statistics, vol. 7, 264-291.
    [ pdf ] • [ journal website ] • [ software ]
  9. B. Guedj and G. Guillot (2011). Estimating the Location and Shape of Hybrid Zones.
    Molecular Ecology Resources, vol. 11(6), 1119-1123.
    [ pdf ] • [ journal website ] • [ software ]

Academic publications

  1. B. Guedj (2013). Agrégation d'estimateurs et de classificateurs : théorie et méthodes. Ph.D. thesis, UPMC.
    [ pdf ] • [ link ]