Guillaume Obozinski's webpage

Address:
Groupe Will
Département d'Informatique
45, rue d'Ulm
Paris Cedex 75230

I just joined WILLOW, a joint research team between INRIA Rocquencourt, École Normale Supérieure de Paris and Centre National de la Recherche Scientifique as a post-doctoral researcher, after completing my PhD in the Statistics department at UC Berkeley.
My PhD advisor at Berkeley was Pr. Michael Jordan.
Before my time in Berkeley I studied Mathematics and Computer Science at the École Normale Supérieure de Cachan, near Paris.

I am interested in Machine Learning and its applications, in particular in Computational Biology.
My current work is related to variable selection, the LASSO and regularization schemes inducing sparsity or dimensionality reduction. In particular I am interested in mechanisms that couple feature selection between several learning problems. I am also interested by online learning, semi-supervised learning and missing data.
I have been working with Pr. William Noble at the University of Seattle on a project on protein function prediction. I am also involved in a long term collaboration with the group of Steven Brenner on a project on automated protein classification. These projects involve large multi-class or structured prediction problems.

An efficient Matlab implementation of the SMO algorithm proposed by Francis Bach and Gert Lanckriet to solve the Support Kernel Machine (SKM) optimization aka Multiple Kernel Learning. ( See Bach, F.R. & Lanckriet, G.R.G., Jordan, M.I. (2004). Fast Kernel Learning using Sequential Minimal Optimization . Technical Report CSD-04-1307, Division of Computer Science, University of California, Berkeley )
The code for Multi-task Feature Selection has been incorporated in the Berkeley Transfert Learning Toolkit (Matlab).

	Guillaume Obozinski Post-doctoral researcher WILLOW Research team, INRIA - ENS, Paris	Address: Groupe Will Département d'Informatique 45, rue d'Ulm Paris Cedex 75230 Office: Haut du DI E-mail:
About me	I just joined WILLOW, a joint research team between INRIA Rocquencourt, École Normale Supérieure de Paris and Centre National de la Recherche Scientifique as a post-doctoral researcher, after completing my PhD in the Statistics department at UC Berkeley. My PhD advisor at Berkeley was Pr. Michael Jordan. Before my time in Berkeley I studied Mathematics and Computer Science at the École Normale Supérieure de Cachan, near Paris.	My Resume: CV [pdf]
Research	I am interested in Machine Learning and its applications, in particular in Computational Biology. My current work is related to variable selection, the LASSO and regularization schemes inducing sparsity or dimensionality reduction. In particular I am interested in mechanisms that couple feature selection between several learning problems. I am also interested by online learning, semi-supervised learning and missing data. I have been working with Pr. William Noble at the University of Seattle on a project on protein function prediction. I am also involved in a long term collaboration with the group of Steven Brenner on a project on automated protein classification. These projects involve large multi-class or structured prediction problems.	.
Publications and Reports	New: Group Lasso with Overlap and Graph Lasso. L. Jacob, G. Obozinski, J.-P. Vert. International Conference on Machine Learning (ICML), to appear, 2009. New: Joint covariate selection and joint subspace selection for multiple classification problems, G. Obozinski, B.Taskar, M.I. Jordan, Statistics and Computing, 2009 (to appear). High-dimensional union support recovery in multivariate regression, G. Obozinski, M.J. Wainwright, M.I. Jordan, Preproceedings of the 21st International Conference on Neural Information Processing Systems (NIPS), 2009. [appendix]. Union support recovery in high-dimensional multivariate regression, G. Obozinski, M.J. Wainwright, M.I. Jordan, Berkeley Stat. Technical Report 761, August 2008. Consistent probabilistic output for protein function prediction, G. Obozinski, C. Grant, G. Lanckriet, M.I. Jordan, W. S. Noble, Genome Biology 2008, 9(Suppl 1):S6. A critical assessment of Mus musculus gene function prediction using integrated genomic evidence, Lourdes Peña-Castillo, Murat Tasan, Chad L Myers, Hyunju Lee, Trupti Joshi, Chao Zhang, Yuanfang Guan, Michele Leone, Andrea Pagnani, Wan Kyu Kim, Chase Krumpelman, Weidong Tian, Guillaume Obozinski, Yanjun Qi, Sara Mostafavi, Guan Ning Lin, Gabriel F Berriz, Francis D Gibbons, Gert Lanckriet, Jian Qiu, Charles Grant, Zafer Barutcuoglu, David P Hill, David Warde-Farley, Chris Grouios, Debajyoti Ray, Judith A Blake, Minghua Deng, Michael I Jordan, William S Noble, Quaid Morris, Judith Klein-Seetharaman, Ziv Bar-Joseph, Ting Chen, Fengzhu Sun, Olga G Troyanskaya, Edward M Marcotte, Dong Xu, Timothy R Hughes, Frederick P Roth, Genome Biology 2008, 9(Suppl 1):S2. Multi-task Feature Selection, G. Obozinski, B. Taskar, M. I. Jordan, UC Berkeley Technical Report [pdf], 2007. Work presented at the ICML-06 Workshop on Structural Knowledge Transfer for Machine Learning, Pittsburgh, PA, 2006. Imaging Brain Activation Streams from Optical Flow Computation on 2-Riemannian Manifolds, J. Lefèvre, G. Obozinski, S. Baillet, IPMI 2007: 470-481.	Coauthors: Laurent Jacob Jean-Philippe Vert Martin J. Wainwright Ben Taskar William S. Noble Michael I. Jordan Gert Lanckriet
Matlab code available	An efficient Matlab implementation of the SMO algorithm proposed by Francis Bach and Gert Lanckriet to solve the Support Kernel Machine (SKM) optimization aka Multiple Kernel Learning. ( See Bach, F.R. & Lanckriet, G.R.G., Jordan, M.I. (2004). Fast Kernel Learning using Sequential Minimal Optimization . Technical Report CSD-04-1307, Division of Computer Science, University of California, Berkeley ) The code for Multi-task Feature Selection has been incorporated in the Berkeley Transfert Learning Toolkit (Matlab).	Download: SKMsmo.tar
Links	Figures in latex Chris Jordan, the "Photography of Large Numbers" Transportation for the future: Solar-powered roller-blading-robot-pulled chariot