Stéphane Canu ,
Laboratoire LITIS EA 4108 - INSA de Rouen
02 32 95 98 44
- Je suis le Directeur du LITIS et animateur du
TAC, le thème apprentissage et contexte au sein du laboratoire
LITIS
- j'enseigne au sein du Département
ASI de l'INSA de Rouen,- j'anime le groupe de lecture da Rouen sur la theorie statistique de l'apprentissage - I belong to the pascal network of excelence, - j'ai passé un an de congé de recherches dans le groupe machine learning de l'ANU à Canberra - je co-organise un workshop a NIPS on Temporal Segmentation - nous mettons à disposition deux boites à outils matlab pour les SVM : SVM and Kernel Methods Matlab Toolbox - |
|
Projets
Enseignements
A few recent slides / Quelques présentations récentes
Publications 2011
Xilan Tian, Gilles Gasso, Stéphane Canu: A Multi-kernel Framework for Inductive Semi-supervised Learning. ESANN 2011
Emilie Niaf, Rémi Flamary, Carole Lartizien, Stéphane Canu: Handling uncertainties in SVM classification CoRR abs/1106.3397: (2011)
Alain Rakotomamonjy, Rémi Flamary, Gilles Gasso, Stéphane Canu: ellp-ellq Penalty for Sparse Linear and Sparse Multiple Kernel Multitask Learning. IEEE Transactions on Neural Networks 22(8): 1307-1320 (2011)
Publications 2010
Stéphane Canu: Recent Advances in Kernel Machines. CIARP 2010: 1
Hachem Kadri, Emmanuel Duflos, Philippe Preux, Stéphane Canu, Manuel Davy: Nonlinear functional regression: a functional RKHS approach. Journal of Machine Learning Research - Proceedings Track 9: 374-380 (2010)
Publications 2009
General Framework for Nonlinear Functional Regression with Reproducing Kernel Hilbert Spaces H Kadri, E Duflos, M Davy, P Preux, S Canu - 2009 - hal.inria.fr
Recovering sparse signals with non-convex penalties and DC programming G Gasso, A Rakotomamonjy, S Canu - 2009 - Accepted for publication
Publications 2008
Y. Grandvalet,J. Keshet, A. Rakotomamonjy, S. Canu: Suppport Vector Machines with a Reject Option Accepted at NIPS
A. Rakotomamonjy, F. Bach, Y. Grandvalet, S. Canu: SimpleMKL, Journal of Machine Learning Research, to appear, 2008.
G. Gasso, A. Rakotomamonjy, S. Canu: Recovering sparse signals with non-convex penalties and DC programming Accepted at MLSP
Karina Zapien Arreola, Thomas Gärtner, Gilles Gasso, Stéphane Canu: Regularization path for Ranking SVM. ESANN 2008: 415-420
New methods for the identification of a stable subspace model for dynamical systems G Mallet, G Gasso, S Canu - IEEE Workshop on Machine Learning for Signal Processing, 2008
Smoothness and sparsity tuning for Semi-Supervised SVM Gilles GASSO, Karma ZAPIEN and Stephane CANU, in Mining Massive Data Sets for Security - Advances in Data Mining, Search, Social Networks and Text Mining, and their Applications to Security Edited by Françoise Fogelman-Soulié, Domenico Perrotta, Jakub Piskorski, Ralf Steinberger, 2008
Publications 2007
Alain Rakotomamonjy, Francis Bach, Stéphane Canu, Yves Grandvalet: More efficiency in multiple kernel learning. ICML 2007: 775-782
Gaëlle Loosli and Léon Bottou and Stéphane Canu Training Invariant Support Vector Machines using Selective Sampling in "Large Scale Kernel Machines" MIT Press",2007
Gaëlle Loosli and Stéphane Canu Comments on the "Core Vector Machines: Fast SVM Training on Very Large Data Sets'' JMLR, 8(Feb):291--301, 2007 pdf
Gaëlle Loosli and Gilles Gasso Stéphane Canu Regularization paths for nu-SVM and nu-SVR ISNN, International Symposium on Neural Networks, 2007
Gaëlle Loosli and Stéphane Canu Programmation quadratique et apprentissage. Grande taille et parcimonie Optimisation en traitement du signal et de l'image", editor "Patrick Siarry",2007
G. Gasso, K. Zapién, S. Canu. L1-Norm Regularization Path for Sparse Semi-Supervised Laplacian SVM. Accepted in ICMLA 2007, Cincinnati, Ohio, USA.
K. Zapién, G. Gasso, S. Canu. Smoothness and L1-Norm Regularization Paths for Semi-Supervised Laplacian SVM. Slovenian Workshop on Graph Theory, Bled 2007.
K. Zapién, G. Gasso, S. Canu. Estimation of Tangent planes for Neighborhood Graph Correction, ESANN April 2007, Brugges.
G. Gasso, K. Zapién, S. Canu. Computing and Stopping the Solution Paths for µ-SVR, ESANN April 2007, Brugges.
Gilles Gasso, Karina Zapien et Stéphane Canu, Chemins de régularisation pour les modèles de régression de type nu-SVR, GRETSI 2007
V. Guigue, A. Rakotomamonjy, S. Canu, Estimation de signaux par noyaux d'ondelettes. to appear in Traitement du Signal, 2007.
Stéphane CANU Machines à noyaux pour l'apprentissage statistique Dossier : TE5255 Date de parution : 02/2007 Techniques de l'ingénieur
Publications 2006
Stéphane Canu and Alex Smola Kernel methods and the exponential family Neurocomputing, Volume 69, Issues 7-9, March 2006, Pages 714-720 pdf
M. Davy, F. Desobry and S. Canu, "Estimation of Minimum Measure Sets in Reproducing Kernel Hilbert Spaces and Applications.", IEEE ICASSP 2006, Toulouse, May 2006. pdf
V. Guigue, A. Rakotomamonjy, S. Canu, Kernel Basis Pursuit, in Revue d'Intelligence Artificielle, New methods in Machine Learning, Vol 20, N°6 , pp 757-774, 2006.
V. Guigue, A. Rakotomamonjy, S. Canu, Translation invariant classification of non-stationary signals Neurocomputing, Vol 69, pp 743-753, 2006.
"G. Loosli and Sang-Goog Lee and V. Guigue and A. Rakotomamonjy and S. Canu ", " Perception d'états affectifs et apprentissage ", "RIA Revue d'intelligence artificielle, Edition spéciale Interactions Emotionnelles ", 2006
Gaëlle Loosli and Stéphane Canu and Léon Bottou, SVM et apprentissage des très grandes bases de données, CAp Conférence d'apprentissage, 2006.
Publications 2005
ESANN'05 - Kernel methods and the exponential family, Canu Stéphane and Smola Alex, National ICT Australia (Australia) pdf et les slides
AMSDA'05 - Invariances in Classification: an efficient SVM implementation : pdf et les slides
CAP 2005 - Context Retrieval by Rupture Detection, Gaëlle Loosli, San-Goog Lee, Stéphane Canu pdf
UM 2005 - Context changes detection by One class SVMs, Gaëlle Loosli, San-Goog Lee, Stéphane Canu pdf
Publications 2004
Slides of the Berder summer school on machine learning:
G. Loosli and S. Canu and S.V.N. Vishwanathan and Alexander J. Smola and Manojit Chattopadhyay, Une boîte à outils rapide et simple pour les SVM, CAp 2004 - Conférence d'Apprentissage, pp 113-128,Presses Universitaires de Grenoble, isbn = 9-782706-112249 (ps english translation : ps)
C.S. Ong, S. Canu "Regularization by Early Stopping", Technical Report, Computer Sciences Laboratory, RSISE, ANU, 2004. (Argues a new way to look at regularization, which looks at filter functions on the spectrum) PS.GZ
C.S. Ong, X. Mary, S.
Canu, A.J. Smola,
"Learning with Non-Positive Kernels", In Proceedings of the 21st
International Conference on Machine Learning, 2004. pp. 639-646.
(What happens when we do not have a positive definite kernel? It turns
out that we can still do learning, in what is now a Krein Space. This
paper talks about regression with indefinite kernels.) pdf
- here is the associated poster
A. Rakotomamonjy, X. Mary, S.
Canu : "Non
Parametric regression with wavelet kernels"; Applied Stochastics Model
for Business and Industry; pp 1-18 (2004) pdf
Publications 2003
X. Mary, D. De Brucq et S. Canu : Sous-dualités et noyaux (reproduisants) associés . Comptes rendus - Mathématiques. Volume 336/11 pp. 949-954, 2003. acces par science direct
Stéphane Canu, Xavier Mary and Alain Rakotomamonjy : Functional learning through kernel. ( version postscript ) (version.pdf )J. Suykens, G. Horvath, S. Basu, C. Micchelli, J. Vandewalle (Eds.) Advances in Learning Theory: Methods, Models and Applications, NATO Science Series III: Computer and Systems Sciences, Vol. 190, pp 89-110, IOS Press, Amsterdam 2003.
Pour le projet sur l'apprentissage du contexte voir les pages de V. Guigue et de G. Loosli
Publications 2002
Yves Grandvalet and Stéphane Canu : Adaptive Scaling for Feature Selection in SVMs, accepted for publication at NIPS 2002. (version postscript , version pdf )Publications 2001Stéphane Canu, Xavier Mary and Alain Rakotomamonjy : Functional learning through kernel.
Slides of the talk given at the NATO Advanced Study Institute on Learning Theory and Practice (LTP 2002) July 11 2002 - K.U. Leuven Belgium, ( slides in pdf )
Stéphane Canu, Alain Rakotomamonjy et Xavier Mary, Construction de noyaux pour les Machines à Vecteur Support (SVM). Application aux noyaux d'ondelettes . Exposé effectué dans le cadre de la Journée Thématique " Retours des sciences cognitives sur l'apprentissage automatique " au Laboratoire Leibniz à Grenoble le 20 Juin 2002 ( sources tex , version postscript , version pdf )
S. Canu : Introduction aux SVM . Séminaire au GREYC à Caen le 17 Janvier 2002 ( sources tex , version postscript, version pdf )
S. Canu : Modèles connexionnistes et machines à vecteurs supports pour la décision . Chapitre de l'ouvrage "Décision de reconnaissance des formes en Signal", éditeur R. Lengellé, Hermès 2002.
A. Rakotomamonjy, S. Canu, Frame Kernels for Learning, accepted for publication in ICANN 2002. ( version pdf )
A. Rakotomamonjy, X. Mary, S. Canu and D. De Brucq, Frame Kernels for Learning, Complete version. Rapport interne PSI 02-005 ( version ps , version pdf )
Associated with this paper we propose an exemple of wavelet kernel: the "Daudechie" waveletKernel (AVI)
and in 1D and in 2dKanevski M., Parkin R., Pozdnukhov A., Timonin V., Maignan M., Yatsalo B., Canu S. (2002) Environmental Data Mining and Modelling Based on Machine Learning Algorithms and Geostatistics. International Environmental Modeling and Software society conference (iEMSs2002), Lugano, Switzerland, pp. 414-419.
Kanevski M., Pozdnoukhov A., Canu S., Maignan M. (2002) Advanced Spatial Data Analysis and Modelling with Support Vector Machines. International Journal of Fuzzy Systems, Vol. 4, No. 1, March 2002, pp. 606-616
S. Canu : Modèles connexionnistes et machines à vecteurs supports pour la décision . Rapport interne PSI 01-003 (version postscript ) (bibliographie .bbl )S. Canu, A. Rakotomamonjy, Ozone peak and pollution forecasting using Support Vectors, IFAC Workshop on environmental modelling, Yokohama, 2001Yokohama, 2001. ( version pdf )
A. Rakotomamonjy, S. Canu, Estimation de la concentration en ozone par SVM, Actes d' Automatique et Environnement 2001, Saint Etienne. ( version pdf )
V. Demyanov, S. Soltani, M. Kanevski, S. Canu, M. Maignan, E. Savelieva, V. Timonin, V. Pisarenko: "Wavelet analysis residual kriging vs. neural network residual kriging" Stochastic Environmental Research and Risk Assessment 15 (2001) 1, 18-32. ( version pdf )
A. Rakotomamonjy, S. Canu, Frame, Reproducing Kernel, Regularization and Learning, Technical Report, Perception Systeme Information, Insa de Rouen, Submitted to Journal of Learning Machine Research, March 2001, Modification January 2002. ( version pdf )
Publications 2000Pozdnoukhov A., Kanevski, M., Demyanov V., Canu S., Maignan M., Kravetski A., Parkin R., Savelieva E., Trutce A., Chernov S. (2001) Environmental Data Mining with Geostatistics and Machine Learning Algorithms, 4-th INTAS Interdisciplinary Symposium on Physical and Chemical Methods in Biology, Medicine and Environment, Moscow, May 30-June 3, 2001.
Kanevski M., Pozdnukhov A., Canu S., Maignan M., Wong P. Shibli S. (2001) Support Vector Machines for Classification and Mapping of Reservoir Data. A chapter from “Soft computing for reservoir characterization and modelling”, Springer-Verlag, 2001, pp. 531-558.
S. Canu and A.
Elisseff, "Regularization, kernels and sigmoid nets",
Unpublished
manuscript (
postscript version of the paper )
Abstract: Many Neural networks use
sigmoid--shaped functions because of biology-inspired arguments.
But there must be some mathematical reason for their
efficiency.
This work aims at presenting such a mathematical reason. It
emerges
when the input probability distribution \mu is taken
into account in the definition of the regularization term in the
framework
of regularization theory for learning. In this case the solution of the regularized minimizing
algorithm
( i.e . the learning machine) depends on a
cumulative
distribution function, i.e. a sigmoid--shaped function.
S. Canu and A Elisseff, "pourqoi les réseaux de neurones de type perceptron multicouche conviennent-'ils a l'appentissage". presnetation lors des journées "Theorie de l'Apprentissage et Modelisation Cognitive" du Pole Rhone-Alpes à lyon les 5 et 6 mai 1999 ( presentation powerpoint )
R. Le Riche, G. Cailletaud et S. Canu, "Vers un non-linéaire matériau automatique", 4ème Colloque en calul des Structures, Giens, France, Mai 1999
Liste des publications 1998 : Laboratoire PSI.
Projets de recherche :
INTAS 1957 : Environmental Data Analysis with Learning Based Algorithms
1997-1999 INTAS Grant 96-1957 (Started in February 1998)
Participants:
Réseaux de neurones et ondelettes pour l'identification des modèles de type boite noire
Participants:
Stéphane Boucheron, Stéphane Canu, Patrick Gallinari et Yves Grandvalet : " De la représentation à la validation : la généralisation dans les réseaux de neurones ". Travail effectué dans le cadre du groupe apprentissage du PRC-IA.
Voici le
même article mais sous la forme d'une page WEB.
Last Modified: 26 Janvier, 2008,
Stéphane Canu (scanu <<AT>> insa-rouen.fr ) 02 32 95 98 44
