Link to COG weekly seminars here (Google Sheets).
Events Search and Views Navigation
January 2024
Seminar: Bayesian Optimization in Molecule Space: Challenges and Opportunities
Title: Bayesian Optimization in Molecule Space: Challenges and Opportunities Speaker: Austin Tripp Affiliation: University of Cambridge Location: Huxley 315 Abstract. Rational design of experiments in chemistry is one of the most commonly mentioned applications of Bayesian optimization (BO). Therefore you might presume that existing BO algorithms for chemistry are well-developed. In this talk I explain how performing BO on the discrete, structured space of molecules introduces extra complexity to BO which standard methods do not handle well. I will outline specific problems…
Find out more »March 2020
Seminar by Prof. Oliver Stein (Cancelled due to the Covid-19 situation)
Title: Pessimistic Bilevel Optimization Pessimistic bilevel optimization problems, as optimistic ones, possess a structure involving three interrelated optimization problems. Moreover, their finite infima are only attained under strong conditions. We address these difficulties within a framework of moderate assumptions and a perturbation approach which allow us to approximate such finite infima arbitrarily well by minimal values of a sequence of solvable single-level problems. To this end, as already done for optimistic problems, we introduce the standard version of the pessimistic…
Find out more »December 2019
Seminar by Associate Professor Jakob Nordström
TITLE: Learn to Relax: Integrating Integer Linear Programming with Conflict-Driven Search ABSTRACT: Pseudo-Boolean (PB) solvers optimize 0-1 integer linear programs by extending the conflict-driven learning paradigm from SAT solving. Though PB solvers should be exponentially more efficient than SAT solvers in theory, in practice they can sometimes get hopelessly stuck even when the relaxed linear program (LP) is infeasible over the reals. Inspired by mixed integer programming (MIP), we address this problem by interleaving incremental LP solving with cut…
Find out more »Seminar by prof. Miguel Anjos
Professor Miguel Anjos from the University of Edinburgh is giving a seminar on: Tight-and-Cheap Conic Relaxations for AC Optimal Power Flow and Optimal Reactive Power Dispatch Abstract: The classical alternating current optimal power flow problem is nonconvex and generally hard to solve. We propose a new conic relaxation obtained by combining semidefinite optimization with RLT. The proposed relaxation is stronger than the second-order cone relaxation, competitive with the recently proposed QC relaxation, and up to one order of magnitude…
Find out more »November 2019
Seminar with Prof. Greg Sorkin: Extremal Cuts and Isoperimetry in Random Cubic Graphs
The minimum bisection width of random cubic graphs is of interest because it is one of the simplest questions imaginable in extremal combinatorics, and also because the minimum bisection of (general) cubic graphs plays a role in the construction of efficient exponential-time algorithms, and it seems likely that random cubic graphs are extremal. It is known that a random cubic graph has a minimum bisection of size at most 1/6 times its order (indeed this is known for all cubic…
Find out more »July 2019
Seminar: Largest Small n-Polygons: Numerical Results and Conjectured Optima
Title: Largest Small n-Polygons: Numerical Results and Conjectured Optima Speaker: János D. Pintér Affiliation: Department of Industrial and Systems Engineering, Lehigh University Location: 218 Huxley Building Time: 14:00 - 15:00 Abstract. LSP(n), the largest small polygon with n vertices, is defined as the polygon of unit diameter that has maximal area A(n). Finding the configuration LSP(n) and the corresponding A(n) for even values n >= 6 is a long-standing challenge that can be also perceived as class of hard global optimization problems. We…
Find out more »Seminar: Chordal Completions – Semidefinite Programming and Minimum Completions
Title: Chordal Completions - Semidefinite Programming and Minimum Completions Speaker: Dr Arvind Raghunathan Affiliation: Mitsubishi Electric Research Laboratories (MERL) Location: 217 Huxley Building Time: 14:00 - 15:00 Abstract. A graph is chordal if every cycle of length at least four contains a chord, that is, an edge connecting two nonconsecutive vertices of the cycle. Chordal completion of a given undirected graph G is a chordal graph, on the same vertex set, that has G as a subgraph. Several classical applications…
Find out more »June 2019
Seminar: Extremal Cuts and Isoperimetry in Random Cubic Graphs
Title: Extremal Cuts and Isoperimetry in Random Cubic Graphs Speaker: Prof. Gregory B. Sorkin Affiliation: Dept of Mathematics, The London School of Economics and Political Science (LSE) Location: LT139 Huxley Building Time: 14:00 - 15:00 Abstract. The minimum bisection width of random cubic graphs is of interest because it is one of the simplest questions imaginable in extremal combinatorics, and also because the minimum bisection of (general) cubic graphs plays a role in the construction of efficient exponential-time algorithms, and it…
Find out more »May 2019
Seminar: Robust Discrete Optimization: Globalized Gamma Robustness and Radius of Robust Feasibility
Title: Robust Discrete Optimization: Globalized Gamma Robustness and Radius of Robust Feasibility Speaker: Prof. Dr Frauke Liers Affiliation: Dept of Mathematics, FAU Erlangen-Nürnberg Location: 217 Huxley Building Time: 15:00 - 16:00 Abstract. In this talk, we extend the notion of two robust optimization methodologies that were originally introduced for continuous problems towards robust discrete tasks. On the one hand, we look at globalized robust optimization that has been proposed as a generalization of the standard robust optimization framework in order to allow for a controlled decrease…
Find out more »Seminar: Exact and heuristic MIP methods for the solution of MINLP – Examples from gas transport optimization problems
Title: Exact and heuristic MIP methods for the solution of MINLP - Examples from gas transport optimization problems Speaker: Dr Lars Schewe Affiliation: Dept of Mathematics, FAU Erlangen-Nürnberg Location: 217 Huxley Building Time: 13:30 - 14:30 Abstract. In this talk, we present exact and heuristic methods for MINLP, the development of which was motivated by applications in gas transport optimization. In this talk, we present a sample of our approaches and focus on provable results for both the exact and the heuristic methods. The…
Find out more »