FREE ELECTRONIC LIBRARY - Dissertations, online materials

Pages:   || 2 |

«B remi.bardenet Rémi Bardenet Í Research interests Adaptive and approximate Markov chain ...»

-- [ Page 1 ] --

Rémi Bardenet – CV and publication list

B remi.bardenet@gmail.com

Rémi Bardenet Í http://www.2020science.net/people/remi-bardenet

Research interests

Adaptive and approximate Markov chain Monte Carlo, Bayesian optimization, Gaus-


sian processes, applications to particle physics and computational biology.

Academic positions

Postdoctoral fellow, Dept. of Statistics, University of Oxford, Oxford, UK.

January 2013-

2020 Science fellowship of the EPSRC, working with Chris Holmes on large-scale Markov chain Monte Carlo methods, motivated by biological data analysis. See track record below for more details.

2020 science (www.2020science.net) is a network of computational scientists creating effective approaches to answer fundamental life science questions. This programme is funded by the British applied science research council (EPSRC).

Education Doctorat (Ph.D.), Université Paris-Sud XI, Orsay (France), très honorable.

2009-2012 Computer science, Towards adaptive learning and inference - Applications to hyperparameter tuning and astroparticle physics, under the supervision of Balázs Kégl (LAL, LRI, CNRS).

Très honorable is the highest honours you can get in Université Paris-Sud.

My thesis was also awarded the 2nd prize for the best French Ph.D. in computer science for the year 2012/2013 (Gilles Kahn prize of the Société Informatique de France).

Please ask me for the reports by my thesis reviewers (Éric Moulines and Christian Robert) and defense jury if you want a detailed external opinion on my work.

Master (M.Sc.), Ecole Normale Supérieure, Cachan (France), summa cum laude.

2008-2009 Mathematics, computer vision and machine learning.

Courses given, e.g., by Gabor Lugosi, Francis Bach, Jean-Yves Audibert, Jean-Philippe Vert, Donald Geman, Rémi Munos, Yali Amit.

Agrégation, Université Louis Pasteur, Strasbourg (France), national rank 82.

2008 Mathematics major, probability & statistics minor.

The agrégation is a national competitive exam, originally meant to hire teachers for undergraduate levels. A large proportion (2800+ candidates that year for 252 positions) of French students in mathematics take it, as it is also considered a sesame to the Ph.D. program.

Magistère, Université Louis Pasteur, Strasbourg (France), summa cum laude.

2005-2009 Applied and fundamental mathematics.

Magistères are selective university programs made of traditional B.Sc./M.Sc. programs with privileges such as small-group exercise sessions and advanced complementary courses.

Licence (B.Sc.), Université Louis Pasteur, Strasbourg (France), summa cum laude.

2005-2006 Applied and fundamental mathematics.

Classe préparatoire, Lycée J.-B. Kléber, Strasbourg (France).

2003-2005 Mathematics & physics major, computer science minor.

Classes préparatoires are advanced and competitive undergraduate programs.

1/6 Rémi Bardenet – CV and publication list Master thesis & notable scholar projects M.Sc. Thesis, Université Paris-Sud XI, Orsay (France).

2009 Sampling-based inference methods for the Pierre Auger experiment (high-energy physics), supervised by Balázs Kégl (LAL, LRI, CNRS).

Internship, Institut de Mathématiques de Jussieu, Paris (France).

2007 2-month internship on Descriptive set theory and its applications to the existence of universal spaces for convexity properties, supervised by Gilles Godefroy.

Dissertation, Université Louis Pasteur, Strasbourg (France).

2006 Dissertation on Gödel’s work on completeness, supervised by Michel Émery. This dissertation won the first prize ex-aequo among the 20 dissertations in the magistère that year.

Internship, CNRS, Strasbourg (France).

2005 Short scholar internship and first contact with particle physics on The measure of a muon’s lifetime, supervised by Abdelnour Lounis.

Teaching Experience Lecturer, Univ. Oxford, Oxford (UK).

2013-2014 8 hours, teaching half of the course Advanced simulation on Monte Carlo methods (4th year students).

Class tutor, Univ. Oxford, Oxford (UK).

2013 14 hours, tutoring for the course Advanced simulation on Monte Carlo methods (4th year students).

Teaching assistant, Univ. Paris-Sud XI, Orsay (France).

2009–2012 64 hrs. per year. Covered topics include Linear and Nonlinear Programming (L3), Stochastic processes (M1), C programming (L2) and orientation of students (L1).

L1, L2, and L3 correspond to the first, second, and third year that lead to a bachelor’s degree, M1 and M2 to the two years of master.

Examiner, Lycée J.-B. Say, Paris (France).

2008–2009 Maths examiner for undergraduate students from “classes préparatoires”.

Private teacher, Strasbourg (France).

2005–2008 Private courses in Mathematics for undergraduate students from “classes préparatoires”.

Grants and awards awarded the 2nd prize for the best French Ph.D. in computer science (Gilles Kahn 2014 prize 2013).

awarded a Junior research fellowship at St Cross college, Oxford.

2013–2016 2020 Science fellowship of the British engineering and physical sciences research coun2015 cil (EPSRC).

Ph.D. scholarship of the University of Paris-Sud XI and the French ministry of ad2012 vanced education and research.

Merit-based scholarship of the French Ministry of National Education (this kind of 2003–2008 scholarship disappeared in 2008).

2/6 Rémi Bardenet – CV and publication list Notable invited talks Invited to give the Markov chain Monte Carlo methods lecture at the joint HelmholtzMax Planck Institut winter school on Monte Carlo Methods, Max Planck Institut Munich.

Invited to give the Numerical Bayesian methods lecture at the SOS statistics summer 2012 school of the IN2P3 (French national institute for particle and nuclear physics). A book chapter resulted from this lecture.

Reviewing activities I have reviewed papers for Journal of Machine Learning Research, Transactions on 2009– Pattern Analysis and Machine Intelligence, and International Conference on Machine Learning.

Supervision I have co-supervised Ahmed Lasmar’s master internship on Modelling the electromagnetic component of the Auger tank signal.

–  –  –

Interests Music I play the piano, keyboards and synthesizers, and I am teaching myself guitar and drums. I am currently especially interested in improvised music.

–  –  –

Track Record Statistical methodology. I have been particularly interested in self-tuning inference and optimization methods, also denoted as adaptive stochastic search methods. Indeed, for today’s complex and large-scale models resulting from modelling natural phenomena, inference has to be automatized to be realistically undertaken. My main contribution so far to statistical methodology was to propose, apply and study an adaptive Metropolis algorithm with online relabeling [19, 5, 2], that performs Markov chain Monte Carlo simulation with adaptive proposals in models with symmetries, such as mixture models. This algorithm is novel in that it is adaptive both in its proposal mechanism and in the definition of the target distribution, which makes its theoretical study both challenging and interesting, combining tools from stochastic approximation and vector quantization. More recently, I have been developing adaptive approximate Bayesian computation methods (ABC) motivated by phylogenetic applications, and MCMC algorithms that rely on subsampling to scale to datasets of billions of datapoints with provable guarantees [3]; see also [1] for technical tools I developed for this second project. On the application side, I have recently started a collaboration with the computational biology group of Microsoft Research Cambridge on model choice.

Machine learning. My contributions in machine learning (ML) are in the subfield of hyperparameter optimization. Still with the idea in mind to deliver turn-key algorithms, I first studied and proposed an improvement to Bayesian optimization algorithms [7], a useful paradigm for hyperparameter tuning. In 2011, I was invited to spend a month with Yoshua Bengio’s group at Université de Montréal, to develop methods that apply to the automatic tuning of deep learning networks [6]. Deep learning algorithms are a class of state-of-the-art ML algorithms which can have tens of hyperparameters, the tuning of which is usually referred to as black magic by nonexperts, thus justifying the development of automatic tuning methods. Motivated by the positive results we obtained in [6], I have been recently working on including information on dataset similarity into automatic hyperparameter tuning algorithms [8].

Particle physics. During my thesis, I was a member of the Pierre Auger collaboration, a team of 300+ researchers and technicians devoted to the construction, maintenance, and data analysis of the Auger cosmic ray observatory. I have thus participated in several internal collaboration conferences, discussed and coauthored a number of journal papers with the full collaboration (see publication list). My direct and personal contributions include probabilistic modelling of the low-level signal [21] and complex empirical Bayes inference procedure [20]. More marginally, I made technical contributions such as radio measurements [22] and spent a month on the site of the observatory in Argentina in 2011 on a shift, monitoring the telescopes of the experiment.

Publications My research is multidisciplinary, and publishing habits differ among the relevant disciplines. I have published so far in computer science, in computational statistics (both abusively denoted by CS), and in physics.

–  –  –

In machine learning, top journals are Journal of Machine Learning Research and Machine Learning. However, it is considered as important and prestigious to publish in the following three conferences: NIPS (Neural and Information Processing Systems), UAI (Uncertainty in Artificial Intelligence), ICML (International Conference in Machine Learning) and to a lesser extent AISTATS (Artificial Intelligence and Statistics).

In statistics, journal papers are more valued than conference papers, and major journals are more numerous. Bernoulli is one of the top journals.

The process of publication in experimental physics is pretty different. Journal papers are more valued, but they often have long author lists. As I was a member of the 300+-member Auger Collaboration, I coauthored journal papers that were the result of the work of the full collaboration. Astroparticle Physics is the main journal in the field of Astroparticles, to which the Auger experiment is dedicated. My main personal contributions to Auger (modelling and inference) are underlined in “short-author-list” conference proceedings and internal papers (GAP notes). The latter are shown here under technical reports, and are not available outside of the Auger Collaboration.

As of November 2013 and according to Google Scholar, my publications in CS have been cited 61 times, and my publications in Physics 550+ times.

CS journals [1] R. Bardenet and O.-A. Maillard. Concentration inequalities for sampling without replacement.

accepted in Bernoulli, available as arxiv.org/abs/1309.4029, 2013.

[2] R. Bardenet, O. Cappé, G. Fort, and B. Kégl. Adaptive MCMC with online relabeling. accepted in Bernoulli, available as arxiv.org/abs/1210.2601, 2013.

CS refereed proceedings [3] R. Bardenet, A. Doucet, and C. Holmes. Towards scaling up MCMC: an adaptive subsampling approach. accepted in Proceedings of the International Conference on Machine Learning (ICML), 2014.

[4] R. Bardenet, M. Brendel, B. Kégl, and M. Sebag. Collaborative hyperparameter tuning. In International Conference on Machine Learning (ICML), 2013.

[5] R. Bardenet, O. Cappé, G. Fort, and B. Kégl. An adaptive Metropolis algorithm with online relabeling. In Proceedings of the International Conference on Artificial Intelligence and Statistics (AISTATS), volume 22, pages 91–99, April 2012.

[6] J. Bergstra, R. Bardenet, B. Kégl, and Y. Bengio. Algorithms for hyperparameter optimization.

In Advances in Neural Information Processing Systems (NIPS), volume 24. The MIT Press, 2011.

[7] R. Bardenet and B. Kégl. Surrogating the surrogate: accelerating Gaussian-process-based global optimization with a mixture cross-entropy algorithm. In Proceedings of the 27th International Conference on Machine Learning (ICML), 2010.

CS refereed workshops [8] R. Bardenet, M. Brendel, B. Kégl, and M. Sebag. SCoT: surrogate-based collaborative tuning for hyperparameter learning that remembers the past. NIPS workshop on Bayesian Optimization and Decision Making, 2012.

[9] R. Bardenet, B. Kégl, and G. Fort. Relabeling MCMC algorithms in Bayesian mixture learning.

Snowbird Learning workshop, 2011.

[10] R. Bardenet and B. Kégl. Sampling-based optimization with mixtures. NIPS workshop on Optimization for Machine Learning, 2009.

–  –  –

CS book chapters

[11] R. Bardenet. Proceedings of the 2012 IN2P3 School of Statistics, chapter Monte Carlo methods.

EDP Sciences, 2013.

Physics Journals (selected papers) [12] P. Abreu et al. The lateral trigger probability function for UHE cosmic rays showers detected by the Pierre Auger Observatory. Astroparticle Physics, 35:266–276, 2011. Auger Collaboration paper.

[13] P. Abreu et al. Anisotropy and chemical composition of ultra-high energy cosmic rays using arrival directions measured by the Pierre Auger Observatory. Journal of Cosmology and Astroparticle Physics, page 022, 2011. Auger Collaboration paper.

[14] P. Abreu et al. Advanced functionality for radio analysis in the Offline software framework of the Pierre Auger Observatory. Nuclear Instruments and Methods in Physics Research A, 635:92–102, 2011. Auger Collaboration paper.

[15] P. Abreu et al. The Pierre Auger Observatory scaler mode for the study of solar activity modulation of galactic cosmic rays. JINST, 6:01003, 2011. Auger Collaboration paper.

[16] P. Abreu et al. Search for first harmonic modulation in the right ascension distribution of cosmic rays detected at the Pierre Auger Observatory. Astroparticle Physics, 34:627–639, 2011. Auger Collaboration paper.

Pages:   || 2 |

Similar works:

«Title: Squamous cell carcinoma of the oral tongue in young non-smokers is genomically similar to tumors in older smokers.Authors: Curtis R Pickering1, Jiexin Zhang2, David M Neskey1, Mei Zhao1, Samar A Jasser1, Jiping Wang1, Alexandra Ward1, C Jillian Tsai3, Marcus V Ortega Alves1, Jane H Zhou4, Jennifer Drummond6, Adel K El-Naggar4, Richard Gibbs6,7, John N Weinstein2,5, David A Wheeler6, Jing Wang2, Mitchell J Frederick1, Jeffrey N Myers1 Affiliations: 1 Department of Head and Neck Surgery,...»

«Development of functional biomaterials by self-assembled nanostructures. Tatsuyuki Yoshii 2014 Preface and Acknowledgements The studies presented in this dissertation have been carried out under the direction of Professor Itaru Hamachi at the Department of Synthetic Chemistry and Biological Chemistry, Graduate School of Engineering, Kyoto University, from April 2005 to September 2014. The study is focused on the development of functional biomaterials by self-assembled nanostructures. My...»

«Available online at www.pelagiaresearchlibrary.com Pelagia Research Library European Journal of Experimental Biology, 2013, 3(6):252-255 ISSN: 2248 –9215 CODEN (USA): EJEBAU Comparison of traditional Doogh (yogurt drinking) and Kashk characteristics (Two traditional Iranian dairy products) Arezoo Noori, Fariba Keshavarzian, Sepideh Mahmoudi*, Morvarid Yousefi and Leila Nateghi Department of Food Science and Technology, Varamin-Pishva Branch, Islamic Azad University, Varamin, Iran _ ABSTRACT...»

«Fisheries Stream Survey Manual Appendix 7. Appendix 7. Introduction to Common Fish Diseases Fishing is a very popular outdoor activity in Minnesota and many fishermen are interested in the welfare of the fish. Consequently, the fish they catch are often scrutinized for anything unusual, and fishery biologists and fish pathologists are questioned for identification of these conditions. Because fish, like humans are attacked by a wide variety of parasites, bacteria, viruses and anomalies, many...»

«Radiative transfer modelling for sun glint correction in marine satellite imagery Submitted by Susan Barbara Kay to the University of Exeter as a thesis for the degree of Doctor of Philosophy in Biological Sciences In December 2011. This thesis is available for Library use on the understanding that it is copyright material and that no quotation from the thesis may be published without proper acknowledgement. I certify that all material in this thesis which is not my own work has been...»

«1 Human Papillomavirus: Biology and Pathogenesis José Veríssimo Fernandes1 and Thales Allyrio Araújo de Medeiros Fernandes2 1Federal University of Rio Grande do Norte 2University of Rio Grande do Norte State Brazil 1. Introduction The human papillomavirus (HPV) is one of the most common causes of sexually transmitted disease in both men and women around the world, especially in developing countries, where the prevalence of asymptomatic infection varies from 2 to 44%, depending on the...»

«The Effect of Thalidomide, an Angiogenesis Inhibitor, on the Estrus Cycle and Reproductive Function of Female Mice Ashley Dockendorff Department of Biology Hartwick College Oneonta, NY This thesis is submitted in partial satisfaction of the requirements for the degree of Bachelor of Arts from the Department of Biology, Hartwick College. _ _ Thesis Advisor Date _ _ Chair, Biology Department Date The Effect of Thalidomide, an Angiogenesis Inhibitor, on the Estrus Cycle and Reproductive Function...»

«International Journal of Comparative Psychology, 2003, 16, 65-84. Copyright 2003 by the International Society for Comparative Psychology Application of Behavioral Knowledge to Conservation in the Giant Panda Ronald R. Swaisgood Zoological Society of San Diego, U.S.A. Xiaoping Zhou, Gwiquan Zhang, Wolong Nature Reserve, China Donald G. Lindburg, Zoological Society of San Diego, U.S.A. and Hemin Zhang Wolong Nature Reserve, China Over the past several years we have developed a research program to...»

«Biotic and abiotic factors influencing diversification of herbivorous mammals by Lucy Tran A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Ecology and Evolutionary Biology) in the University of Michigan 2015 Doctoral Committee: Professor L. Lacey Knowles, Chair Research Associate Professor Liliana Cortés Ortiz Professor Diarmaid Ó Foighil Professor John C. Mitani © Lucy Tran 2015 To my parents, for allowing me to follow my dreams,...»

«Digestive Systems: The Anatomy of Representative Vertebrates Modified from: Biology in the laboratory. 3rd edition. Helms, Helms, Kosinski and Cummings. Biological Investigations: Form, Function, Diversity and Process. 7th Edition. W.D. Dolphin Helms, Helms, Kosinski, Cummings. Biology in the Laboratory, 3rd edition. Freeman Publishing. Harold M. Kaplan and Kathleen A. Jones Southern Illinois University OVERVIEW The digestive system participates in the procurement and metabolism of...»

«Case 4:04-cv-02688-JEJ Document 309 Filed 11/04/2005 Page 1 of 18 Appendix A: Excerpts from Of Pandas and People (2nd ed., 1993), the published version used by students, Quote A: “This book has a single goal: to present data from six areas of science that bear on the central question of biological origins. We don't propose to give final answers, nor to unveil The Truth. Our purpose, rather, is to help readers understand origins better, and to see why the data may be viewed in more than one...»

«Local Self-Renewing of Microglia after Genetic Ablation is Dependent on Interleukin-1 Signaling Dissertation Zur Erlangung des Grades Doktor der Naturwissenschaften Am Fachbereich Biologie Der Johannes Gutenberg-Universität Mainz Vorgelegt von Julia Bruttger geb. am 14. Juni 1984 in Bad Kreuznach, Deutschland Mainz, 2015 Tag der mündlichen Prüfung: 25.01.2016_ Für Sven Table of Contents Table of Contents Table of Contents Abbreviations 1 Introduction 1.1 Microglia – The Link between...»

<<  HOME   |    CONTACTS
2016 www.dissertation.xlibx.info - Dissertations, online materials

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.