YURI BOYKOV Department of Computer Science Cornell University, Ithaca, NY 14853 yura@cs.cornell.edu http://www.cs.cornell.edu/home/yura (607) 255-1165 work (607) 257-8530 home EDUCATION: - Cornell University, Ithaca, NY, 1996. Ph.D. in Operations Research with emphasis in Applied Probability and Statistics - Cornell University, Ithaca, NY, 1995. M.S. in Operations Research, minors in Optimization and Finance - Moscow Institute of Physics and Technology (MIPT), Russia, 1992. M.S. in Electrical Engineering and Cybernetics (with honors) RESEARCH ACTIVITIES: - GENERAL INTERESTS, Keywords: stochastic modeling, optimization, algorithms, dynamic programming, foundations of statistics, optimal control, computer simulation - CURRENT RESEARCH, Keywords: feature based object recognition, tracking, learning object models, visual correspondence, stereo vision, motion, optical flows, image segmentation - POST-DOCTORAL RESEARCH, Computer Science, Cornell: My post-doctoral research is concerned with efficient energy minimization algorithms applicable in computer vision. I developed a technique based on graph cuts that finds a strong local minimum of a wide class of energy functions whose global minimization is known to be NP hard. This method gives promising experimental results in stereo vision, motion estimation, and image restoration. I also apply the same graph cut technique to MAP estimation of a match between an object and an image in an effective Bayesian recognition framework. As a special case of this framework I derived a simple Hausdorff- Kalman tracking algorithm that allows real time tracking in highly cluttered scenes when the object is partially or (in some frames) completely occluded. A possibility for adaptive object representations that can be learned in the process of tracking is investigated. - PH.D. DISSERTATION, Operations Research, Cornell: My dissertation ``Two problems in Statistics and Mathematical Finance'' is devoted to the previously unresolved problem of characterizing coherent inferences in the countably additive setting. The main idea is to ask a statistician to accept bets on subsets of a parameter space at prices based on his own inference. The inference is coherent if we cannot ``beat'' the statistician. My main result says that an inference function is coherent if and only if it arises from a proper countably additive Bayesian prior. The proof is drawn from the arbitrage theorems of mathematical finance based on a clear financial interpretation of coherence; the existence of a Bayesian prior for an inference function can be seen as an analog for existence of a martingale measure. - M.S. THESIS, Moscow Institute of Physics and Technology: My thesis ``Stochastic dynamic programming in the problem of optimal pursuit'' (1992) is based on my contributions to a joint project on simulating a space system proposed in the Strategic Defense Initiative. The goal was to use theoretically optimal components/algorithms. In my work I use dynamic programming to derive and implement an optimal control algorithm for the stochastic problem of pursuit. REFEREED JOURNAL PUBLICATIONS: - Y. Boykov, O. Veksler, R. Zabih A Variable Window Approach to Early Vision In IEEE Transactions on Pattern Analysis Machine Intelligence, vol. 20, no. 12, pp. 1283-1294, 1998 REFEREED CONFERENCE PUBLICATIONS: - Y. Boykov, D. Huttenlocher A New Bayesian Framework for Object Recognition In IEEE Computer Vision and Pattern Recognition Conference, 1999, to appear. - Y. Boykov, O. Veksler, R. Zabih Markov Random Fields with Efficient Approximations In IEEE Conf. Computer Vision & Pattern Recognition, pp. 648-655, 1998 - Y. Boykov, O. Veksler, R. Zabih Disparity Component Matching for Visual Correspondence In IEEE Conf. Computer Vision & Pattern Recognition, pp. 470-475, 1997 - Y. Boykov Stochastic dynamic programming in the optimal pursuit problem In Discrete and Numerical Methods of Signal Processing (Conf. at Moscow Inst. of Physics and Technology), 1992. SUBMITTED PAPERS: - Y. Boykov, O. Veksler, R. Zabih Energy Minimization with Discontinuities Submitted to The International Journal on Computer Vision (1998). - Y. Boykov Arbitrage theorem for statistical inference systems Submitted to the Annals of Statistics (1997). - Y. Boykov, D. Huttenlocher A Bayesian Framework for Model Based Tracking Submitted to the International Conference on Computer Vision (1999). - Y. Boykov, O. Veksler, R. Zabih Fast Approximate Energy Minimization via Graph Cuts Submitted to the International Conference on Computer Vision (1999). - Y. Boykov, O. Veksler, R. Zabih A New Algorithm for Energy Minimization with Discontinuities Submitted to the International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, (1999). OTHER PAPERS: - Y. Boykov, D. Huttenlocher A New Bayesian Framework for Object Recognition In DARPA Image Understanding Workshop, vol. 1, pp. 259-266, 1998. - Y. Boykov, O. Veksler, R. Zabih Efficient Restoration of Multicolor Images with Independent Noise Tech. Rep. Document ID ncstrl.cornell/TR98-1712, Cornell University, Computer Science, 1998. - Y. Boykov, P. Carr Method of Lines Evaluation of American Options with Constant Dividends and Asian Options, Research paper, 1996. WORK EXPERIENCE: - Fall 1997 - Present Cornell University, Ithaca, NY Post-doctoral research assistant, Department of Computer Science Developing and testing efficient algorithms for visual correspondence, object recognition, and tracking based on Bayesian models. - Fall 1996 - Summer 1997 Carnegie Mellon University, Pittsburgh, PA Visiting Assistant Professor, Department of Mathematics - Fall 1990 - Spring 1992 Moscow Institute of Physics and Technology, Russia Research assistant Developing and implementing optimal stochastic pursuit algorithms for a computer model of the space system proposed by the Strategic Defense Initiative. - programming experience in C, Pascal, MatLab HONORS AND AWARDS: - McMullen Fellowship from Cornell University - Diploma with honors from the Moscow Institute of Physics and Technology REFERENCES: - Prof. Ramin Zabih: Cornell, CS, (607)-255-8413, rdz@cs.cornell.edu - Prof. Dan Huttenlocher: Cornell, CS, (607)-255-3036, dph@cs.cornell.edu - Prof. Eva Tardos: Cornell, CS, (607)-255-0984, eva@cs.cornell.edu - Prof. Jon Kleinberg: Cornell, CS, (607)-255-3600, kleinber@cs.cornell.edu - Prof. Steven Shreve: CMU, Math.Dep., (412)-268-8484, shreve@andrew.cmu.edu - Dr. Peter Carr, Vice-President: Morgan Stanley, (212)-761-7340, carrp@morgan.com