Research Interests

I am interested in Machine Learning and Optimization. In particular, in extending and combining novel algorithmic techniques from the areas of reinforcement learning, combinatorial and convex optimization, and dynamical systems, so as to design efficient algorithms with provable guarantees that approximately solve challenging real-world problems.

---

Publications

• Learning Diverse Skills for Local Navigation under Multi-constraint Optimality
  • Project web site
  IEEE International Conference on Robotics and Automation (ICRA 2024)
  Jin Cheng, Marin Vlastelica, Pavel Kolev, Chenhao Li, Georg Martius


• Diverse Offline Imitation Learning
  • Project web site
  Preprint 2023
  Marin Vlastelica, Jin Cheng, Georg Martius, Pavel Kolev


• On Imitation in Mean-field Games
  Advances in Neural Information Processing Systems 36 (NeurIPS 2023)
  
  • Poster and Video
  Giorgia Ramponi, Pavel Kolev, Olivier Pietquin, Niao He, Mathieu Laurière, Matthieu Geist


• Online Learning under Adversarial Nonlinear Constraints
  Advances in Neural Information Processing Systems 36 (NeurIPS 2023)
  
  • Poster and Video
  Pavel Kolev, Georg Martius, Michael Muehlebach
  Invited talk at ETH AI Center 2023 and INSAIT 2023


• Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions
  • Project web site
  IEEE International Conference on Robotics and Automation (ICRA 2023)
  Chenhao Li, Sebastian Blaes, Pavel Kolev, Marin Vlastelica, Jonas Frey, Georg Martius


• Benchmarking Offline Reinforcement Learning on Real-Robot Hardware
  • Project web site
  International Conference on Learning Representations (ICLR 2023)
  Nico Gürtler, Sebastian Blaes, Felix Widmaier, Manuel Wüthrich, Pavel Kolev, Stefan Bauer, Bernhard Schölkopf, Georg Martius


• Secretary and Online Matching Problems with Machine Learned Advice
  Journal of Discrete Optimization 2023
  A preliminary version of this paper was presented at the Advances in Neural Information Processing Systems 33 (NeurIPS 2020)
  
  • The full arXiv version
  Antonios Antoniadis, Themis Gouleakis, Pieter Kleer, Pavel Kolev


• Physarum Multi-Commodity Flow Dynamics
  Journal of Theoretical Computer Science (TCS), Section C (Theory of Natural Computing 2022)
  
  • The full arXiv version
  Vincenzo Bonifaci, Enrico Facca, Frederic Folz, Andreas Karrenbauer, Pavel Kolev, Kurt Mehlhorn, Giovanna Morigi, Golnoosh Shahkarami, Quentin Vermande


• Convergence of the Non-Uniform Directed Physarum Model
  Journal of Theoretical Computer Science (TCS), Section C (Theory of Natural Computing 2020)
  
  • The full arXiv version
  Enrico Facca, Andreas Karrenbauer, Pavel Kolev, Kurt Mehlhorn


• Convergence of the Non-Uniform Physarum Dynamics
  Journal of Theoretical Computer Science (TCS), Section C (Theory of Natural Computing 2020)
  
  • The full arXiv version
  Andreas Karrenbauer, Pavel Kolev, Kurt Mehlhorn


• A PTAS for $\ell_{0}$-Low Rank Approximation
  Annual ACM-SIAM Symposium on Discrete Algorithms (SODA 2019)
  
  • The full arXiv version
  Frank Ban, Vijay Bhattiprolu, Karl Bringmann, Pavel Kolev, Euiwoong Lee, David P. Woodruff


• Approximation Algorithms for $\ell_{0}$-Low Rank Approximation
  Advances in Neural Information Processing Systems 30 (NeurIPS 2017)
  
  • The full arXiv version and Poster
  Karl Bringmann, Pavel Kolev, David P. Woodruff


• Two Results on Slime Mold Computations
  Journal of Theoretical Computer Science (TCS), Section C (Theory of Natural Computing 2019)
  A preliminatry version was presented at the International Conference on Operations Research (OR 2017)
  
  • The full arXiv version and Slides
  Ruben Becker, Vincenzo Bonifaci, Andreas Karrenbauer, Pavel Kolev, Kurt Mehlhorn


• Density Independent Algorithms for Sparsifying k-Step Random Walks
  International Workshop on Approximation Algorithms for Combinatorial Optimization Problems (APPROX/RANDOM 2017)
  
  • The full arXiv version and Slides
  Gorav Jindal, Pavel Kolev, Richard Peng, Saurabh Sawlani


• Approximate Spectral Clustering: Efficiency and Guarantees
  A preliminary version was presented at the Annual European Symposium on Algorithms (ESA 2016) and
  (as a contributed talk) at the IGAFIT Highlights of Algorithms conference (HALG 2016)
  
  • Slides
  Pavel Kolev, Kurt Mehlhorn
  


• An Efficient Parallel Algorithm for Spectral Sparsification of Laplacian and SDDM Matrix Polynomials
  Gorav Jindal, Pavel Kolev


• Dirichlet Eigenvalues, Local Random Walks, and Analyzing Clusters in Graphs
  International Symposium on Algorithms and Computation (ISAAC 2014)
  Pavel Kolev, He Sun

---

Teaching

• Teaching Assistant at Eberhard Karls University of Tübingen
• Summer 2023: Reinforcement Learning, given by Prof. Georg Martius


• Teaching Assistant at Saarland University
• Summer 2019: Optimization, given by Dr. Andreas Karrenbauer
• Summer 2017: Optimization, given by Dr. Andreas Karrenbauer
• Summer 2016: Randomized Algorithms and Probabilistic Analysis of Algorithms, given by Prof. Kurt Mehlhorn and Prof. Thomas Kesselheim
• Summer 2014: Great Ideas in Theoretical Computer Science, given by Dr. He Sun
• Summer 2013: Optimization, given by Dr. Andreas Karrenbauer and Prof. Matthias Mnich
• Winter 2012: Machine Learning, given by Prof. Matthias Hein

---

Workshops

• Mini-Workshop on Reinforcement Learning (MWRL 2024): January 12, 2024, Mannheim, Germany
• The 16th European Workshop on Reinforcement Learning (EWRL 2023): September 14-16, 2023, Brussels, Belgium
• The 15th European Workshop on Reinforcement Learning (EWRL 2022): September 19-21, 2022, Milano, Italy
• 49th Annual ACM Symposium on the Theory of Computing: June 19 - 23, 2017, Montreal, Canada
• ACM-SIAM Symposium on Discrete Algorithms: January 10 – 29, 2016, Arlington, Virginia, USA
• MADALGO Summer School on Streaming Algorithms: August 10 – 13, 2015, Aarhus University, Denmark
• Conference on Learning Theory: July 3 - 6, 2015, University Pierre and Marie Curie, Paris, France
• Algorithmic Spectral Graph Theory Boot Camp: August 26 – 29, 2014, Simons Institute for the Theory of Computing, UC Berkeley, USA
• ADFOCS Current Trends in Network Optimization: August 11 - 15, 2014, Saarbrücken, Germany
• EADS Summer School on Hashing: Theory and Applications: July 14 – 17, 2014, University of Copenhagen, Denmark
• ADFOCS Current Trends in Theoretical Computer Science: August 13 - 17, 2012, Saarbrücken, Germany

---

Reviewing

• RA-L 2023, ICRA 2023, ICLR 2023, L4DC 2022, Math. Program. 2020, Optim. Lett. 2019
• STOC 2019, JMIV 2018, ITCS 2018, ESA 2017, ICALP 2017, STACS 2017, MFCS 2016

---

Education

• January 2014 - November 2018
  Doctoral degree in Computer Science from the Computer Science Department at Saarland University
  Thesis: Algorithmic Results for Clustering and Refined Physarum Analysis (supervisors: Prof. Dr. Kurt Mehlhorn and Dr. Karl Bringmann)


• October 2011 - November 2013
  Master degree in Computer Science from the Computer Science Department at Saarland University
  Thesis: Community Analysis Using Local Random Walks (supervisors: Prof. Dr. Kurt Mehlhorn and Dr. He Sun)


• September 2007 - July 2011:
  Bachelor degree in Informatics from the Faculty of Mathematics and Informatics at Sofia University

---

Seminar Talks

• Gradient Descent only Converges to Minimizers: ATML-SS17
• Faster spectral sparsification and numerical algorithms for SDD matrices: RGA-SS17
• Approximate Undirected Maximum Flows in O(m polylog(n)) Time: RGA-WS16
• Faster Online Matrix-Vector Multiplication: RGA-SS16
• Streaming Lower Bounds for Approximating MAX-CUT: ALB-WS15
• Spectral Sparsification and Regret Minimization Beyond Matrix Multiplicative Updates: RGA-WS15
• Spectral Sparsification of Random-Walk Matrix Polynomials: RGA-SS15
• Multi-way spectral partitioning and higher-order Cheeger inequalities: RGA-WS12
• Conflict Packing Yields Linear Vertex-Kernels for k-FAST, k-dense RTI and a Related Problem: RGA-SS12

---

---