Matthias Poloczek

Assistant Professor

Educational Background

Doctor of Philosophy, Computer Science, Goethe-University Frankfurt am Main, 2013

Research & Teaching Interests

Design and analysis of algorithms at the interface of machine learning and optimization. Specifically, algorithms that are fast in practice and provide provable performance guarantees. Optimal learning research to optimally trade-off exploration vs. exploitation for sequential decision-making under uncertainty. This research in particular deals with development of Bayesian optimization algorithms for simulation optimization and the design of experiments with applications in aerospace engineering, biochemistry, chemical engineering, materials science and medicine.

Other research includes investigation of the approximation algorithms for combinatorial problems that allow very fast implementations and thus scale to large data, and inherent limitations of the greedy paradigm, i.e., the price of making decisions myopically.

Awards and Honors

  • Alexander von Humboldt-Stiftung, Feodor Lynen Research Fellowship for Postdoctoral Researchers
  • Deutscher Akademischer Austauschdienst, Research Grant for Postdoctoral Researchers, Declined in favor of the Feodor Lynen fellowship
  • Award for outstanding academic performance and for the best PhD in computer science, Goethe-Universität Frankfurt am Main, 2013
  • Award for outstanding academic performance and the best diploma in computer science, Goethe-Universität Frankfurt am Main, 2008

Professional Societies and Activities

  • Program Committee, the Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18), February 2–7 2018, New Orleans, Louisiana
  • Co-organizer, SIAM minisymposium on “Multi-fidelity and Multi-information Source Methods,” 2017 SIAM Conference on Computational Science and Engineering, Feb. 27 to March 3, 2017, Atlanta, Georgia
  • Humboldt Fellow
  • Member of INFORMS and Deutscher Hochschulverband (DHV)
  • External reviewer for journals, including SIAM Journal on Computing (SICOMP), SIAM Journal on Control and Optimization (SICON), Journal of Discrete Algorithms, AIAA Journal, ACS Journal of Chemical Theory and Computation, etc.
  • External reviewer for conferences in machine learning and theoretical computer science, e.g., NIPS, SODA, ESA, ALENEX, etc.

Selected Publications

Bayesian Optimization of Combinatorial Structures.
With Ricardo Baptista.
To appear in the proceedings of the Thirty-fifth International Conference on Machine Learning (ICML), 2018.

Multi-Information Source Optimization.
With Peter I. Frazier and Jialei Wang.
In Proc. of the Thirty-first Annual Conference on Neural Information Processing Systems (NIPS), 2017.
Awarded a spotlight presentation (top 3.5%)

Bayesian Optimization with Gradients.
With Peter I. Frazier, Andrew G. Wilson, and Jian Wu.
In Proc. of the Thirty-first Annual Conference on NIPS, 2017.
Awarded an oral presentation (top 1.5%)

Greedy Algorithms for MAX SAT: Simple Algorithms and Inapproximability Bounds.
With Georg Schnitger, David P. Williamson, and Anke van Zuylen.
SIAM Journal on Computing (SICOMP), 46(3), pp. 1029–1061, 2017.

An Experimental Evaluation of Fast Approximation Algorithms for MAX SAT.
With David P. Williamson.
Journal of Experimental Algorithmics (JEA), 22(1), 2017.

University of Arizona College of Engineering