Past Events

Title: Generalized maximum entropy estimation Speaker: Dr David Sutter Affiliation: Institute for Theoretical Physics, ETH Zurich Location: 217 Huxley Building Time: 14:00 - 15:00 Abstract. We consider the problem of estimating a probability distribution that maximizes the entropy while satisfying a finite number of moment constraints, possibly corrupted by noise. Based on duality of convex programming, we present a novel approximation scheme using a smoothed fast gradient method that is equipped with explicit bounds on the approximation error. This is joint…

Title: Model-based design of optimal experiments using exact confidence regions Speaker: Dr Radoslav Paulen Affiliation: Institute of Information Engineering, Automation and Mathematics, Slovak University of Technology Location: 218 Huxley Building Time: 16:00 - 17:00 Abstract. This talk discusses a model-based optimal experiment design (OED) for nonlinear systems. OED represents a methodology for optimizing the geometry of the parametric joint-confidence regions, which are obtained in an a posteriori analysis of the parameter estimates. The optimal design is achieved by using the available…

Title: Capacity upper bounds for deletion-type channels Speaker: Dr Mahdi Cheraghchi Affiliation: Dept of Computing, Imperial College London Location: 218 Huxley Building Time: 16:00 - 17:00 Abstract. We develop a systematic approach, based on convex programming and real analysis, for obtaining upper bounds on the capacity of the binary deletion channel and, more generally, channels with i.i.d. insertions and deletions. Other than the classical deletion channel, we give a special attention to the Poisson-repeat channel introduced by Mitzenmacher and Drinea (IEEE…

Title: Arrangements of Circles and Spheres leading to Convex Hulls with Minimal Boundaries Speaker: Dr Josef Kallrath Affiliation: Advanced Business Analytics, BASF SE, Germany Location: 218 Huxley Building Time: 15:00 - 16:00 Abstract. We present and solve a new computational geometry optimization problem in 2D and 3D and provide theoretical insights. Circles (2D) and Spheres (3D) with given radii should be arranged in such a way that a) they do not overlap and b) the length of the perimeter or surface…

Title: Random projections in mathematical programming Speaker: Dr Leo Liberti Affiliation: CNRS LIX, École Polytechnique Location: 218 Huxley Building Time: 15:00 - 16:00 Abstract. In the algorithmic trade-off between generality and efficiency, sometimes the only way out is to accept approximate methods. If all else fails, we can always fall back on heuristic methods. But some form of approximation guarantee is usually preferable. In this talk we shall discuss a set of approximating reformulations to various classes of mathematical programming problems…

Title: More Virtuous Smoothing and Embracing the Ghost of Rolle Speaker: Prof Jon Lee Affiliation: Industrial and Operations Engineering, College of Engineering, University of Michigan Location: LT 308 Huxley Building Time: 14:00 - 15:00 Abstract. In the context of global optimization of mixed-integer nonlinear optimization formulations, we consider smoothing univariate concave increasing functions that have poorly behaved derivative at 0 (for example, root functions). Extending earlier work of Lee and Skipper, we give general conditions under which our smoothing is concave…

Title: Computing Pessimistic Leader-Follower Equilibria with Multiple Followers Speaker: Dr Stefano Coniglio Affiliation: Dept. of Mathematical Sciences, Southampton University Location: Room 217 Huxley Building Time: 14:00 - 15:00 Abstract. We investigate the problem of computing a Leader-Follower equilibrium in Stackelberg games where two or more followers react to the strategy chosen by the (single) leader by playing a Nash Equilibrium. We consider two natural cases, the optimistic one where the followers select a Nash Equilibrium maximizing the leader's utility and the pessimistic one…

Title: Large neighbourhood Benders' search Speaker: Dr. Stephen Maher Affiliation: Lancaster University Management School Location: Room 218 Huxley Building Time: 14:00 - 15:30 Abstract. Enhancements for the Benders' decomposition algorithm can be derived from large neighbourhood search (LNS) heuristics. While mixed-integer programming (MIP) solvers are endowed with an array of LNS heuristics, their use is typically limited in bespoke Benders' decomposition implementations. To date, only ad hoc approaches have been developed to enhance the Benders' decomposition algorithm using large neighbourhood search techniques---namely…