Contact

Stefanie Jegelka

Address: Spemannstr. 38
72076 Tübingen
Room number: 204
Phone: +49 7071 601 559
Fax: +49 7071 601 552
E-Mail: stefanie.jegelka
Print page    
Picture of Jegelka, Stefanie

Stefanie Jegelka

Position: PhD Student  Unit: Alumni Schölkopf

My research is concerend with discrete optimization problems in machine learning. Most recently, I have been working on submodular functions and graph cuts.

I have also worked on theoretical aspects of clustering and density estimation, from the perspective of approximation as well as learning theory.

Below is a summary of projects, and links to code. I have also co-organized the NIPS workshop on Discrete Optimization in Machine Learning.

 

Structured problems with submodular cost

The generic setting here is as follows: take a combinatorial optimization problem, and replace the usual sum-of-weights cost by a nonlinear, submodular function. This makes the problem much harder but has nice applications. In particular, we worked on this setting with graph cuts, for which there are applications in inference and computer vision.
In addition, we derived online algorithms for structured decision problems (e.g., online spanning tree, or s-t cut) with submodular cost (in contrast, almost all previous results are for linear costs).
During our work, we came across several functions that are actually not submodular (but maybe some people hoped them to be), and thus I am collecting those functions in a bag of submodular non-examples.

 

Clustering & Graph Cuts

Nearest Neighbor Clustering: we developed a generic algorithm to minimize popular clustering objective functions, such as Normalized Cut or k-means. This method is consistent, thanks to pooling neighboring points. (with Ulrike von Luxburg, Sebastien Bubeck, Michael Kaufmann).
Download Demo Code

Generalized Clustering via Kernel Embeddings: The concept of clustering can be generalized to finding distributions that are maximally separated. The standard case is to measure separation by means of the distribution (k-means). MMD, the maximum mean discrepancy in a feature space, allows to also take higher-order moments into account. As a result, it is possible to e.g. separate two Gaussians with the same mean but different variance. Interestingly, maximizing MMD is closely related to kernel k-means and various other clustering criteria. (with Ulrike von Luxburg, Bernhard Schoelkopf, Arthur Gretton, Bharath Sriperumbudur)

Approximation Algorithms for Tensor Clustering: We prove an approximation factor for tensor clustering (i.e., "cutting a big cube into little cubes") for maybe the simplest possible algorithm. Various divergences are possible, such as Euclidean distance, Bregman divergences etc. (with Suvrit Sra, Arindam Banerjee)

Solution Stability in Linear Programming Relaxations: Graph Partitioning and Unsupervised Learning. Data is often noisy. If there are several good solutions to an optimization problem (here, graph partitioning), then the noise can be the determining part to make one solution optimal. However, how much do we want to trust this solution, if a small perturbation in the data suddenly makes another solution the best? this work proposes a method to test the stability of an optimal graph cut solution to perturbation of the edge weights (i.e., noise).We also show that several common clustering and graph partitioning objectives fall in our framework. (with Sebastian Nowozin)

 

ICA

Fast Kernel ICA: We developed a Newton-like method for kernel ICA that uses HSIC as the independence criterion (with Hao Shen, Arthur Gretton).
Download Code for Fast kernel ICA


Preferences: 
References per page: Year: Medium:

  
Show abstracts

Articles (2):

Shen H Person, Jegelka S Person and Gretton A Person (2009) Fast Kernel-Based Independent Component Analysis IEEE Transactions on Signal Processing 57(9) 3498-3511.
pdf
Jegelka S Person, Bednar JA and Miikkulainen R (2006) Prenatal development of ocular dominance and orientation maps in a self-organizing model of V1 Neurocomputing 69(10-12) 1291-1296.

Conference papers (14):

Jegelka S Person, Bach F and Sra S Person (2013) Reflection methods for user-friendly submodular optimization In: Advances in Neural Information Processing Systems 26, (Ed) C.J.C. Burges, L. Bottou, M. Welling, Z. Ghahramani and K.Q. Weinberger, 27th Annual Conference on Neural Information Processing Systems (NIPS 2013), 1313--1321.
Jegelka S Person and Bilmes J Person (2011) Approximation Bounds for Inference using Cooperative Cut (Ed) Getoor, L. , T. Scheffer, 28th International Conference on Machine Learning (ICML 2011), International Machine Learning Society, Madison, WI, USA, 577-584.
pdf
Jegelka S Person and Bilmes J Person (2011) Online submodular minimization for combinatorial structures (Ed) Getoor, L. , T. Scheffer, 28th International Conference on Machine Learning (ICML 2011), International Machine Learning Society, Madison, WI, USA, 345-352.
pdfpdf
Jegelka S Person and Bilmes J Person (2011) Multi-label cooperative cuts CVPR 2011 Workshop on Inference in Graphical Models with Structured Potentials, 1-4.
pdf
Jegelka S Person and Bilmes J Person (2011) Submodularity beyond submodular energies: coupling edges in graph cuts IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011), IEEE, Piscataway, NJ, USA, 1897-1904.
pdf
Jegelka S Person, Lin H and Bilmes J Person (2011) On Fast Approximate Submodular Minimization In: Advances in Neural Information Processing Systems 24, (Ed) J Shawe-Taylor, RS Zemel, P Bartlett, F Pereira and KQ Weinberger, Twenty-Fifth Annual Conference on Neural Information Processing Systems (NIPS 2011), 460-468.
pdf
Jegelka S Person and Bilmes J Person (2010) Online algorithms for submodular minimization with combinatorial constraints NIPS 2010 Workshop on Discrete Optimization in Machine Learning: Structures, Algorithms and Applications (DISCML), 1-6.
pdf
Jegelka S Person and Bilmes J Person (2009) Notes on Graph Cuts with Submodular Edge Weights NIPS 2009 Workshop on Discrete Optimization in Machine Learning: Submodularity, Sparsity & Polyhedra (DISCML), 1-6.
pdf
Jegelka S Person, Sra S Person and Banerjee A (2009) Approximation Algorithms for Tensor Clustering In: ALT 2009, (Ed) Gavalda, R. , G. Lugosi, T. Zeugmann, S. Zilles, The 20th International Conference on Algorithmic Learning Theory, Springer, Berlin, Germany, 368-383.
pdf
Jegelka S Person, Gretton A Person, Schölkopf B Person, Sriperumbudur BK Person and von Luxburg U Person (2009) Generalized Clustering via Kernel Embeddings In: KI 2009: AI and Automation, Lecture Notes in Computer Science, Vol. 5803, (Ed) B Mertsching, M Hund and Z Aziz, 32nd Annual Conference on Artificial Intelligence (KI), Springer, Berlin, Germany, 144-152.
pdfpdf
Nowozin S Person and Jegelka S Person (2009) Solution Stability in Linear Programming Relaxations: Graph Partitioning and Unsupervised Learning In: ICML 2009, (Ed) Danyluk, A. , L. Bottou, M. Littman, 26th International Conference on Machine Learning, ACM Press, New York, NY, USA, 769-776.
pdf
von Luxburg U Person, Bubeck S Person, Jegelka S Person and Kaufmann M (2008) Consistent Minimization of Clustering Objective Functions In: Advances in neural information processing systems 20, (Ed) Platt, J. C., D. Koller, Y. Singer, S. Roweis, Twenty-First Annual Conference on Neural Information Processing Systems (NIPS 2007), Curran, Red Hook, NY, USA, 961-968.
pdf
Shen H Person, Jegelka S Person and Gretton A Person (2007) Fast Kernel ICA using an Approximate Newton Method In: JMLR Workshop and Conference Proceedings Volume 2: AISTATS 2007, (Ed) Meila, M. , X. Shen, 11th International Conference on Artificial Intelligence and Statistics, MIT Press, Cambridge, MA, USA, 476-483.
pdf
Jegelka S Person, Gretton A Person and Achlioptas D Person (2005) Kernel ICA for Large Scale Problems NIPS 2005 Workshop on Large Scale Kernel Machines, -.

Contributions to books (1):

Jegelka S Person and Gretton A Person: Brisk Kernel ICA, 225-250. In: Large Scale Kernel Machines, (Ed) Bottou, L. , O. Chapelle, D. DeCoste, J. Weston, MIT Press, Cambridge, MA, USA, (2007).
pdf

Technical reports (5):

Jegelka S Person and Bilmes J Person: Cooperative Cuts for Image Segmentation, UWEETR-1020-0003, University of Washington, Washington, DC, USA, (2010).
Jegelka S Person and Bilmes J Person: Cooperative Cuts: Graph Cuts with Submodular Edge Weights, 189, Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, (2010).
pdf
Sra S Person, Jegelka S Person and Banerjee A : Approximation Algorithms for Bregman Clustering Co-clustering and Tensor Clustering, 177, Max-Planck Institute for Biological Cybernetics, Tübingen, Germany, (2008).
pdf
Kulis B , Sra S Person and Jegelka S Person: Scalable Semidefinite Programming using Convex Perturbations, TR-07-47, University of Texas, Austin, TX, USA, (2007).
Shen H Person, Jegelka S Person and Gretton A Person: Geometric Analysis of Hilbert Schmidt Independence criterion based ICA contrast function, PA006080, National ICT Australia, Canberra, Australia, (2006).

Theses (2):

Jegelka S Person: Combinatorial Problems with Submodular Coupling in Machine Learning and Computer Vision, ETH Zürich, Switzerland, (2012). PhD thesis
Jegelka S Person: Statistical Learning Theory Approaches to Clustering, Eberhard-Karls-Universität Tübingen, Tübingen, Germany, (2007). Diplom thesis
pdf

Talks (3):

Jegelka S Person (2011): Cooperative Cuts: a new use of submodularity in image segmentation, Second I.S.T. Austria Symposium on Computer Vision and Machine Learning, Klosterneuburg, Austria.
Jegelka S Person (2011): Cooperative Cuts, COSA Workshop: Combinatorial Optimization, Statistics, and Applications, München, Germany.
Jegelka S Person and Bilmes J Person (2010): Cooperative Cuts: Graph Cuts with Submodular Edge Weights, 24th European Conference on Operational Research (EURO XXIV), Lisboa, Portugal.
pdf

Export as:
BibTeX, XML, Pubman, Edoc