Sebastian Nowozin
Welcome!
My name is Sebastian Nowozin and I am researcher at Microsoft Research, Cambridge, UK. On this page you find publications and software related to my work.
News
- November 2011. Hard discrete energy minimization instances, resulting from Decision Tree Fields, are available for download.
October 2011. PhD
scholarships in the area of machine learning and high-level computer
vision available, as part of a joint program between the Max-Planck society
and Microsoft Research Cambridge.
More details here.
- June 2011. The CVPR tutorial on structured learning and prediction are online now.
Research
My main research interest is in developing machine learning techniques suitable for solving high-level computer vision tasks, such as image classification and object recognition.
High-level computer vision tasks are a unique source of hard machine learning problems for three reasons. First, in contrast to physics-based processes we do not know the correct model (model uncertainty). Second, humans excel at all high-level vision tasks and thus can provide data and assess model performance (ground truth oracle). Third, image and video data is available for free at an enormous scale (data availability). These properties make computer vision a particularly attractive area for machine learning research.
I am particularly interested in using mathematical optimization as a tool to solve computer vision machine learning tasks.
Publications and related materials
2011
- Sungwoong Kim, Sebastian Nowozin, Pushmeet Kohli, Chang D. Yoo, "Higher-Order Correlation Clustering for Image Segmentation", (PDF), 25th Annual Conference on Neural Information Processing Systems (NIPS 2011).
- Sebastian Nowozin, Carsten Rother, Shai Bagon, Toby Sharp, Bangpeng Yao, Pushmeet Kohli, "Decision Tree Fields", (PDF, supplementary materials, talk slides, poster, hard energy minimization instances (142MB)), 13th International Conference on Computer Vision (ICCV 2011).
Suvrit Sra,
Sebastian Nowozin, and
Stephen J. Wright (Editors),
"Optimization for Machine Learning",
(publisher
link),
edited volume to be published December 2011, Neural
Information Processing series, MIT
Press.
- Patrick Pletscher, Sebastian Nowozin, Pushmeet Kohli, Carsten Rother, "Putting MAP back on the map", (PDF, supplementary), 33rd Annual Symposium of the German Association for Pattern Recognition (DAGM 2011).
Tutorial: Sebastian Nowozin and
Christoph H. Lampert,
"Structured Learning and Prediction in Computer Vision",
(PDF, updated May 2011),
Foundations and
Trends in Computer Graphics and Vision series
of now publishers.
Tutorial slides online.
- Dhruv Batra, Sebastian Nowozin, Pushmeet Kohli, "Tighter Relaxations for MAP-MRF Inference: A Local Primal-Dual Gap based Separation Algorithm", (PDF), International Conference on Artificial Intelligence and Statistics (AISTATS 2011).
- Taesup Kim, Sebastian Nowozin, Pushmeet Kohli, Chang D. Yoo, "Variable Grouping for Energy Minimization", (PDF), IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011).
- Hiroto Saigo, Andre Altmann, Jasmina Bogojeska, Fabian Müller, Sebastian Nowozin, Thomas Lengauer, "Learning from Past Treatments and Their Outcome Improves Prediction of In Vivo Response to Anti-HIV Therapy", (PDF), Statistical Applications in Genetics and Molecular Biology, Vol. 10 (2011), Issue 1.
2010
- Sebastian Nowozin, Peter V. Gehler, and Christoph H. Lampert, "On Parameter Learning in CRF-based Approaches to Object Class Image Segmentation", (PDF, supplementary), 11th European Conference on Computer Vision (ECCV 2010).
- Tutorial: Carsten Rother and Sebastian Nowozin, "Higher-order Models in Computer Vision", (homepage, PDF slides), IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010).
- Sebastian Nowozin and Christoph H. Lampert, "Global Interactions in Random Field Models: A Potential Function Ensuring Connectedness", (PDF, publisher link), SIAM Journal on Imaging Sciences (SIIMS), Vol. 3, Issue 4, 2010.
2009
Sebastian Nowozin,
"Learning with Structured Data: Applications to Computer Vision",
(PDF)
PhD
dissertation at the Technical
University of Berlin.
- Peter V. Gehler and Sebastian Nowozin, "On Feature Combination Methods for Multiclass Object Classification", (PDF, talk slides, project), IEEE International Conference on Computer Vision (ICCV 2009).
- Paramveer S. Dhillon, Sebastian Nowozin, and Christoph H. Lampert, "Combining Appearance and Motion for Human Action Classification in Videos", (PDF), 1st International Workshop on Visual Scene Understanding (ViSU 09).
- Sebastian Nowozin and Stefanie Jegelka, "Solution Stability in Linear Programming Relaxations: Graph Partitioning and Unsupervised Learning", (PDF), International Conference on Machine Learning (ICML 2009).
- Sebastian Nowozin and Christoph H. Lampert, "Global Connectivity Potentials for Random Field Models", (PDF, additional material, project), IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2009).
- Peter V. Gehler and Sebastian Nowozin, "Let the Kernel Figure it Out; Principled Learning of Pre-processing for Kernel Classifiers", (PDF, project) IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2009).
2008
- Sebastian Nowozin and Koji Tsuda, "Frequent Subgraph Retrieval in Geometric Graph Databases", (PDF, project) IEEE International Conference on Data Mining (ICDM 2008).
- Peter V. Gehler and Sebastian Nowozin, "Infinite Kernel Learning", (PDF, project), Max Planck Institute for Biological Cybernetics Techreport TR-178.
- Hiroto Saigo, Sebastian Nowozin, Tadashi Kadowaki, Taku Kudo and Koji Tsuda, "gBoost: A Mathematical Programming Approach to Graph Classification and Regression", (PDF, project), Machine Learning Journal, Springer, Vol 75, Number 1.
- Sebastian Nowozin and Gökhan BakIr, "A Decoupled Approach to Exemplar-based Unsupervised Learning", (PDF, project), 25th International Conference on Machine Learning (ICML 2008).
- Paramveer S. Dhillon, Sebastian Nowozin, and Christoph H. Lampert, "Combining Appearance and Motion for Human Action Classification in Videos", (PDF), Max Planck Institute for Biological Cybernetics Techreport TR-174.
- Sebastian Nowozin and Koji Tsuda, "Frequent Subgraph Retrieval in Geometric Graph Databases", (PDF, project), Max Planck Institute for Biological Cybernetics Techreport TR-180, extended version of ICDM 2008 paper.
2007
- Sebastian Nowozin, Gökhan BakIr, and Koji Tsuda, "Discriminative Subsequence Mining for Action Classification", (PDF, project), IEEE International Conference on Computer Vision (ICCV 2007).
- Sebastian Nowozin, Koji Tsuda, Takeaki Uno, Taku Kudo, and Gökhan BakIr, "Weighted Substructure Mining for Image Analysis", (PDF, additional material, project), IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2007).
Software
- Tuwo - C++ computer vision library
- gboost - graph mining and classification
- pboost - sequence mining and classification
- freqgeo - geometric subgraph mining
- infex - exemplar-based models for unsupervised learning
Archive
- TU-Berlin university homepage (from 2001-2006)
Contact
You can reach me by email at nowozin@gmail.com.