Laurent Candillier
Senior Data Scientist
French born in 1978
lcandillier@hotmail.com

         

Quick presentation

Global cursus

Research works

The global topic I work on is Machine Learning. I first joined the GRAppA laboratory of Lille University and conducted a PhD thesis on Data Mining in collaboration with the company Pertinence. I then joined the Profiling and Data Mining team in Orange Labs to construct an expertise on Recommender Systems. Information Retrieval and Natural Language Processing then completed my centers of interest as director of the R&D team of Nomao. Qualified scientific and technical expert in that areas of research, I then followed up young and innovative companies in the elaboration of their R&D projects and the study and development of their intelligent algorithms. I am now working as Senior Data Scientist in Ocado Technology on projects like Fraud Detection or Demand Forecasting.

Technical report n°28 presents my first research work on automatic learning of user profiles, that falls into the domain of Collaborative Filtering. We demonstrated that such an approach is relevant to help targeting relevant content on news websites. We then stated the interest of combining it to the complementary approach of content-based profiling. Unfortunately, an internal problem at the Rosebud company led to the end of this research.

My works then headed to the domain of Clustering. Inside the company Pertinence, we searched a method that would provide understandable results to the users. Our first contribution, named Tuareg and published under n°26, found its inspiration on the decision trees in supervised learning, aiming at adapting it to unsupervised learning. However, since dividing a dataset by considering the features independently from one another is sometimes not sufficient, we proposed another method called SSC, based on the use of statistical models and published under the n°12 and 25. This method was then adapted to face data having tree formats, leading to very efficient results on large scale experiments, presented on papers n°4 and 15. It demonstrated its strength to face noisy data and provide understandable results. We then extended the method in order to add a hard selection of the most relevant features during learning. That work was presented in paper n°23. Finally, we tackled the important open issue of the evaluation of Clustering algorithms. The new method we have proposed, called cascade evaluation (papers n°11, 14 and 24), allows us to evaluate the interest of the Clustering more objectively than existing methods. All these works led to a PhD thesis untitled Contextualisation, Visualisation and Evaluation in Unsupervised Learning, presented the 15th of september 2006 at Lille University and published under report n°27.

I then made a post-doctorate at Orange Labs on Recommender Systems. We first worked on the study of state-of-the-art approaches and carried out experiments on the main Collaborative Filtering methods that led to the publication of paper n°10. The next paper n°9 shows the interest of developing specific similarity measures for sparse datasets such as those handled in the field of Collaborative Filtering. Finally, paper n°3 offers a large view on the subject, also dealing with the important issue of human-machine interactions (publication n°22). These works led to the delivery of a very efficient and generic engine.

Then nominated director of the Research & Development team of Nomao, my centers of interest enlarged to the topics of Information Retrieval and Natural Language Processing. Paper n°21 gives an overview of the scientific issues raised by the development of Nomao. Publications n°1, 18 and 19 then deal more precisely with our works on Recommender Systems and Information Retrieval, Machine Learning and Natural Language Processing. Important works were conducted on automatically extracting tones from comments collected on the web. In collaboration with the Institute of Research in Computer Science of Toulouse, we co-leaded a PhD thesis on Learning-to-rank in the framework of geolocalization and personalization (publication n°20), and another one about Information Movement Detection. This second study, conducted with Overblog, led us to investigate on the complex issue of searching for diversity in Recommender Systems (publications n°2, 8 and 17). In collaboration with Orange Labs, we organized a workshop and challenge held at the international conference ECML-PKDD 2012: ALRA: Active Learning in Real-world Applications (publication n°13).

As scientific and technical expert certified by the Ministry of Higher Education and Research in France, I then followed up young and innovative companies like TokTokTok, Digimind, Target2Sell, TheFamily, Charly.io, SonetIN, Refactor, France Consultants, Overblog or Ebuzzing - Teads in the elaboration of their R&D projects and the study and development of their intelligent algorithms. As part of our work at TokTokTok, we published the paper n°16 that deals with the interaction between Information Retrieval Systems and Natural Language Processing.

I am now working as Senior Data Scientist in Ocado Technology. Automatic fraud detection, demand forecasting at global or customer level, design of a recommender system or a search engine in Big Data environment are the main topics we deal with in the e-commerce department. Our work on real-time Machine Learning on the Google Cloud Platform has been presented at QCon London and Google Next 2018 conferences (n°5, 6 and 7).

Publications

    Journals: 4

  1. [2014] Systèmes de recommandation et Recherche d'Information
    Laurent Candillier, Étienne Chai, Estelle Delpech
    In Ghislaine Chartron, Imad Saleh, Gérald Kembellec, éditeurs
    Les systèmes de recommandation -- Hermès Sciences, collection information, hypermédias et communication
    [ PDF ] [ BibTeX ]

  2. [2012] Multiple similarities for diversity in recommender systems
    Laurent Candillier, Max Chevalier, Damien Dudognon, Josiane Mothe
    In International Journal on Advances in Intelligent Systems, Volume 5, Number 3 & 4
    [ PDF ] [ BibTeX ]

  3. [2009] State-of-the-Art Recommender Systems
    Laurent Candillier, Kris Jack, Françoise Fessant, Frank Meyer
    In Collaborative and Social Information Retrieval and Access: Techniques for Improved User Modeling, Chapter 1
    [ PDF ] [ BibTeX ]

  4. [2007] Mining XML Documents
    L. Candillier, L. Denoyer, P. Gallinari, M.C. Rousset, A. Termier, A.M. Vercoustre
    In Data Mining Patterns: new Methods and Applications, Chapter 8
    [ PDF ] [ BibTeX ]

    International Conferences: 8

  5. [2018] Machine Learning with Scikit-Learn and Xgboost on Google Cloud Platform
    Laurent Candillier, David Cournapeau, Steve Greenberg
    Google Cloud Next'2018, San Francisco, 24-26 july 2018
    [ Link ] [ Video ]

  6. [2018] Machine Learning with Ease: How Ocado is Building Smart Systems with the Help of GCP
    Laurent Candillier, Przemyslaw Pastuszka
    Google Cloud Next'2018, San Francisco, 24-26 july 2018
    [ Link ] [ Video ]

  7. [2018] Real-Time Decisions Using ML on the Google Cloud Platform
    Laurent Candillier, Carlos Garcia, Jose Jimenez, Przemyslaw Pastuszka
    Distributed Stateful Systems track at Qcon London'2018, UK, 5-9 march 2018
    [ Link ] [ Article ]

  8. [2011] Diversity in Recommender Systems: Bridging the gap between users and systems
    Laurent Candillier, Max Chevalier, Damien Dudognon, Josiane Mothe
    4th International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies and Services
    CENTRIC'2011, Barcelona, Spain, 23-29 october 2011
    [ PDF ] [ BibTeX ] [ Best Paper ]

  9. [2008] Designing Specific Weighted Similarity Measures to Improve Collaborative Filtering Systems
    Laurent Candillier, Frank Meyer, Françoise Fessant
    In Petra Perner, editor
    8th Industrial Conference on Data Mining
    ICDM'2008, Leipzig, Germany, 16-18 july 2008
    Lecture Notes in Computer Science, LNAI 5077, pages 242-255
    [ PDF ] [ Slides ] [ BibTeX ]

  10. [2007] Comparing state-of-the-art collaborative filtering systems
    Laurent Candillier, Frank Meyer, Marc Boullé
    In Petra Perner, editor
    5th International Conference on Machine Learning and Data Mining in Pattern Recognition
    MLDM'2007, Leipzig, Germany, 18-20 july 2007
    Lecture Notes in Computer Science, LNAI 4571, pages 548-562
    [ PDF ] [ Slides ] [ BibTeX ]

  11. [2006] Cascade Evaluation of Clustering Algorithms
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    In Johannes Furnkranz, Tobias Scheffer and Myra Spiliopoulou, editors
    17th European Conference on Machine Learning
    ECML'2006, Berlin, Germany, 18-22 september 2006
    Lecture Notes in Computer Science, LNAI 4212, pages 574-581
    [ PDF ] [ Poster ] [ BibTex ]

  12. [2005] SSC: Statistical Subspace Clustering
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    In Petra Perner and Atsushi Imiya, editors
    4th International Conference on Machine Learning and Data Mining in Pattern Recognition
    MLDM'2005, Leipzig, Germany, 9-11 july 2005
    Lecture Notes in Computer Science, LNAI 3587, pages 100-109
    [ PDF ] [ Slides ] [ BibTeX ]

    International Workshops: 3

  13. [2012] Design and Analysis of the Nomao Challenge - Active Learning in the Real-World
    Laurent Candillier, Vincent Lemaire
    Workshop on Active Learning in Real-world Applications
    ECML-PKDD'2012, Bristol, UK, 28 september 2012
    [ PDF ] [ Slides ] [ BibTeX ]

  14. [2005] Cascade Evaluation of Clustering Algorithms
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    Workshop on Theoretical Foundations of Clustering
    NIPS'2005, Vancouver, Canada, 5-10 december 2005
    [ PDF ] [ Slides ] [ BibTeX ]

  15. [2005] Transforming XML trees for efficient classification and clustering
    Laurent Candillier, Isabelle Tellier, Fabien Torre
    Workshop on Mining XML documents
    INEX'2005, Schloss Dagstuhl, Wadern, Germany, 28-30 november 2005
    [ PDF ] [ Slides ] [ BibTeX ] [ Results ]

    French Conferences: 11

  16. [2016] RI-TAL: le TAL au service de la RI
    Laurent Candillier, Julien Hénot
    13ème Conférence en Recherche d'Information et Applications
    CORIA'2016, Toulouse, 9-11 march 2016
    [ PDF ] [ Slides ] [ BibTex ]

  17. [2013] Diversité de recommandations: application à une plateforme de blogs et évaluation
    Laurent Candillier, Max Chevalier, Damien Dudognon, Josiane Mothe
    10ème Conférence en Recherche d'Information et Applications
    CORIA'2013, Neuchâtel, 3-5 april 2013
    [ PDF ] [ BibTex ]

  18. [2013] Identification de compatibilités entre descripteurs de lieux et apprentissage automatique
    Estelle Delpech, Laurent Candillier, Léa Laporte, Samuel Phan
    13ème Conférence Internationale Francophone sur l'Extraction et la Gestion des Connaissances
    EGC'2013, Toulouse, 29-31 january 2013
    Revue des Nouvelles Technologies de l'Information (RNTI), pages 311-316
    [ PDF ] [ BibTeX ]

  19. [2012] Nomao: un moteur de recherche géolocalisé spécialisé dans la recommandation de lieux et l'e-réputation
    Estelle Delpech, Laurent Candillier
    19ème conférence sur le Traitement Automatique des Langues Naturelles
    TALN'2012, Grenoble, 4-8 june 2012
    [ PDF ] [ BibTeX ]

  20. [2012] Évaluation de la pertinence dans les moteurs de recherche géoréférencés
    Léa Laporte, Laurent Candillier, Sébastien Déjean, Josiane Mothe
    Informatique des Organisations et Systèmes d'Information et de Décision
    INFORSID'2012, Montpellier, 29-31 may 2012
    [ PDF ] [ BibTeX ]

  21. [2011] Nomao: la recherche géolocalisée personnalisée
    Laurent Candillier
    In Djamel A. Zighed and Gilles Venturini, editors
    11ème Conférence Internationale Francophone sur l'Extraction et la Gestion des Connaissances
    EGC'2011, Brest, 25-28 january 2011
    Revue des Nouvelles Technologies de l'Information (RNTI), volume 1, pages 259-261
    [ PDF ] [ Slides ] [ BibTeX ]

  22. [2008] Investigating the Effects of The Types of Feedback in Recommendation Systems
    Kris Jack, Liv Lefebvre, ACK. Laurent Candillier, Frank Meyer
    Conférence sur l'Interaction Homme-Machine
    IHM'2008, Metz, 2-5 septembre 2008
    [ PDF ]

  23. [2006] SuSE: Subspace Selection embedded in an EM algorithm
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    In Laurent Miclet, editor, pages 331-345
    8ème Conférence francophone sur l'Apprentissage automatique
    CAp'2006, Trégastel, 22-24 may 2006
    [ PDF ] [ Slides ] [ BibTeX ]

  24. [2006] Évaluation en cascade d'algorithmes de clustering
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    In Laurent Miclet, editor, pages 109-124
    8ème Conférence francophone sur l'Apprentissage automatique
    CAp'2006, Trégastel, 22-24 may 2006
    [ PDF ] [ Slides ] [ BibTeX ]

  25. [2005] SSC: Statistical Subspace Clustering
    Laurent Candillier, Isabelle Tellier, Fabien Torre, Olivier Bousquet
    In Suzanne Pinson and Nicole Vincent, editors
    5èmes journées francophones d'Extraction et Gestion des Connaissances
    EGC'2005, Paris, 19-21 january 2005
    Revue des Nouvelles Technologies de l'Information (RNTI), volume 1, pages 177-182
    [ PDF ] [ Slides ] [ BibTeX ]

  26. [2004] Tuareg: Classification non supervisée contextualisée
    Laurent Candillier, Isabelle Tellier, Fabien Torre
    In Michel Liquière and Marc Sebban, editors, pages 159-174
    6ème Conférence francophone sur l'Apprentissage automatique
    CAp'2004, Montpellier, 14-16 june 2004
    [ PDF.fr ] [ PDF.eng ] [ Slides ] [ BibTeX ]

    Technical Reports: 2

  27. [2006] Contextualisation, Visualisation et Évaluation en Apprentissage Non Supervisé
    Laurent Candillier
    PhD thesis, Université Charles de Gaulle de Lille 3, September 2006
    [ PDF ] [ Slides ] [ BibTeX ] [ Reports ]

  28. [2001] Apprentissage Automatique de Profils de Lecteurs
    Laurent Candillier, Isabelle Tellier, Fabien Torre
    Technical report GRAppA 2001
    [ PDF ] [ BibTeX ]

Universitary teaching

Computer Science skills

Spoken languages

Other professional experience

Associative life

Leisure