NIPS 2009 Workshop :: Understanding Multiple Kernel Learning Methods

A workshop at the Twenty-Third Annual Conference on Neural Information Processing Systems (NIPS 2009)  7:30am - 6:30pm Friday, December 11, 2009 Hilton: Sutcliffe B   Description Multiple kernel learning has been the subject of nearly a decade of research. Designing and integrating kernels has proven to be an appealing approach to address several, challenging real world applications, involving multiple, heterogeneous data sources in computer vision, bioinformatics, audio processing problems, etc. The goal of this workshop is to step back and evaluate the achievements of multiple kernel learning in the past decade, covering a variety of applications.   In short, this workshop seeks to understand where and how kernel learning is relevant (with respect to accuracy, interpretability, feature selection, etc.), rather than exploring the latest optimization techniques and extension formulations.  More specifically, the workshop envisions to discuss the following two questions:
  1. Kernel learning vs. kernel design: Does kernel learning offer a practical advantage over the manual design of kernels?
  2. Given a set of kernels, what is the optimal way, if any, to combine them (sums, products, learned or non learned, with or without cross-validation)?
We are seeking participants interested in presenting their work and relating their experience in the workshop, providing insight on the above two questions. This includes evidence of MKL improving accuracy beyond any existing method based on single kernels (provided with insights as to why there is such improvement), as well as evidence of the opposite (with insights as to why). We welcome presentation of recent results, as well as presentations based on previously published work that shed light on the above questions.    If you are interested in participating and contributing a presentation, please send the organizers an email with an abstract or a summary. If the presentation is based on previously published work, please include details of such publications.    Repository In conjunction with the workshop, we are establishing an open repository of data sets for use with MKL algorithms.  Authors are encouraged to contribute data to the MKL Repository (mkl.ucsd.edu), and use the repository to benchmark new algorithms.   Schedule
Time Title Authors Materials
7:30--7:45 Multiple kernel learning introduction and workshop goals Gert Lanckriet  
7:45--8:15 Formulations and basic methods and theory Francis Bach and Nathan Srebro slides
8:15--8:45 Learning kernels - theory and survey Mehryar Mohri slides
8:45--9:00 Preliminary analysis of multiple kernel learning: flat maxima, diversity, and Fisher information Theodoros Damoulas, Mark Girolami and Simon Rogers slides
9:00--9:15 Discussion    
9:15--9:50 Coffee break    
9:50--10:20 Designing and combining kernels: some lessons learned from bioinformatics Jean-Philippe Vert slides
10:20--10:30 Poster spotlights    
  Fold recognition using convex combinations of multiple kernels Huzefa Rangwala  
  Multiple kernel learning on imbalanced data: creating a receptor-ligand classifier Ernesto Iacucci, Shi Yu, Fabian Ojeda and Yves Moreau  
  Kernel-based inductive transfer Ulrich Ruckert slides, poster
  Comparing Sparse and Non-sparse Multiple Kernel Learning Marius Kloft, Ulf Brefeld, Soeren Sonnenburg, and Alexander Zien slides, poster
10:30--15:30 Break    
15:30--16:00 Multiple kernel learning for feature selection Manik Varma slides
16:00--16:30 Multiple lernel learning approaches for image classification Peter Gehler  slides
16:30--16:45 Discussion: MKL in vision    
16:45--16:55 Poster spotlights    
  On the algorithmics and applications of a mixed-norm based kernel learning formulation Saketha Nath Jagarlapudi, Dinesh Govindaraj, Raman S, Chiranjib Bhattacharyya, Aharon Ben-Tal and K. R. Ramakrishnan slides, project site
  Localized multiple kernel machines for image recognition Mehmet Gonen and Ethem Alpaydin project site
  Comparison of sparse and nonsparse multiple kernel methods on VOC2009 challenge data Alexander Binder and Motoaki Kawanabe slides
16:55--17:25 Coffee break    
17:25--17:40 Sparsity-accuracy trade-off in MKL Ryota Tomioka and Taiji Suzuki slides, tech report
17:40--18:30 Panel Discussion    
  Posters 
Title Authors
Fold recognition using convex combinations of multiple kernels Huzefa Rangwala
Kernel-based inductive transfer Ulrich Ruckert
An attention-based approach for learning how to fuse decisions of local experts Maryam S. Mirian, Majid Nili Ahmadabadi, Babak N. Araabi and Mohammed H. Zokaei A.
On the Algorithmics and Applications of a Mixed-norm based Kernel Learning Formulation Saketha Nath Jagarlapudi, Dinesh Govindaraj, Raman S, Chiranjib Bhattacharyya, Aharon Ben-Tal, K. R. Ramakrishnan
Localized multiple kernel machines for image recognition Mehmet Gonen and Ethem Alpaydin
Multiple kernel learning on imbalanced data: creating a receptor-ligand classifier Ernesto Iacucci, Shi Yu, Fabian Ojeda and Yves Moreau
Detecting anomalies in multivariate data sets with switching sequences and continuous streams Santanu Das, Bryan Matthews, Kanishka Bhaduri, Nikunj Oza and Ashok Srivastava
Multi-kernel gaussian processes Arman Melkumyan and Fabio Ramos
Comparison of sparse and nonsparse multiple kernel methods on VOC2009 challenge data Alexander Binder and Motoaki Kawanabe
Comparing Sparse and Non-sparse Multiple Kernel Learning
Marius Kloft, Ulf Brefeld, Soeren Sonnenburg, and Alexander Zien
Contextual Object Localization with Multiple Kernel Nearest Neighbor Carolina Galleguillos, Brian McFee, Serge Belongie, Gert Lanckriet
  Organizers