Departement Mathematik

Optimization seminar

×

Modal title

Modal content

Herbstsemester 2012

Datum / Zeit Referent:in Titel Ort
8. Oktober 2012
16:30-18:00 		Prof. Dr. Mirjam Dür
Universität Trier, Deutschland
Details

Optimization Seminar

Titel Copositive optimization, the copositive cone, and its relation with Parrilo's approximations
Referent:in, Affiliation Prof. Dr. Mirjam Dür, Universität Trier, Deutschland
Datum, Zeit 8. Oktober 2012, 16:30-18:00
Ort HG G 19.1
Abstract A copositive optimization problem is a linear problem in matrix variables with the additional constraint that the variable is in the cone of copositive matrices. Here a matrix A is called copositive if the quadratic form (x^T)Ax is nonnegative for all nonnegative x. Copositive optimization problems are of interest, because many quadratic as well as combinatorial problems can be written in this form. A well studied example is the maximum clique problem from graph theory. However, the structure of the copositive cone is not well understood, and in general copositive problems are not tractable. For this reason, a number of relaxations have been proposed, among them a hierarchy of approximating cones was proposed by Parrilo. The talk will give an introduction to these topics and place particular emphasis on the relation between the cone and Parrilo's approximations.
Copositive optimization, the copositive cone, and its relation with Parrilo's approximationsread_more
HG G 19.1
15. Oktober 2012
16:30-18:00 		Prof. Dr. Klaus Schittkowski
Universität Bayreuth, Bayreuth, Deutschland
Details

Optimization Seminar

Titel A New Trust-Region-SQP Algorithm for the Efficient Solution of Non-Convex, Non-Relaxable Mixed-Integer Nonlinear Programming Problems
Referent:in, Affiliation Prof. Dr. Klaus Schittkowski, Universität Bayreuth, Bayreuth, Deutschland
Datum, Zeit 15. Oktober 2012, 16:30-18:00
Ort HG G 19.1
Abstract A new sequential quadratic programming (SQP) algorithm stabilized by trust-regions for solving nonlinear, non-convex and non-relaxable mixed-integer optimization problems is introduced. The mixed-integer quadratic programming subproblems are solved by a branch-and-cut algorithm. Second order information is updated by a modified quasi-Newton update formula (BFGS) applied to the Lagrange function for continuous, but also for integer variables. The design goal is to solve practical optimization problems with expensive executions of an underlying simulation program. Thus, the number of simulations or function evaluations, respectively, is our main performance criterion to measure the efficiency of the code. Numerical results are presented for a set of 186 mixed-integer test problems and different parameter settings of MISQP. The average total number of function evaluations of the new mixed-integer SQP code is less than 1,500 including those needed to approximate partial derivatives.
A New Trust-Region-SQP Algorithm for the Efficient Solution of Non-Convex, Non-Relaxable Mixed-Integer Nonlinear Programming Problemsread_more
HG G 19.1
22. Oktober 2012
16:30-18:00 		Prof. Dr. Joydeep Dutta
Indian Institute of Technology Kanpur, India
Details

Optimization Seminar

Titel Gap Functions and Error Bounds for Variational Inequalities
Referent:in, Affiliation Prof. Dr. Joydeep Dutta, Indian Institute of Technology Kanpur, India
Datum, Zeit 22. Oktober 2012, 16:30-18:00
Ort HG G 19.1
Abstract In this talk we focus on the variational inequalities on finite dimensions. In finite dimensions variational inequalities can be thought of as generalizations of the necessary optimality conditions for minimizing a convex function over a closed convex set. If we consider the convex function to be differentiable then it gives rise to the well studied Stampacchia variational inequality. Further if the convex function is not differentiable then it gives rise to so called generalized variational inequalities or variational inequalities with set-valued maps. It is natural from the point of view of computations to find an error bound to solution set of some well-behaved class of variational inequalities namely monotone and strongly monotone class of variational inequalities. This is usually achieved using the gap function. We motivate as to why we need to use the gap function to develop error bounds. We present some important class of gap functions and their properties which make them relevant in devising error bounds. At the end of the talk we discuss the recent research in the study of gap functions and their applications to error bounds for variational inequalities with set-valued maps and we shall end the talk with an open question.
Gap Functions and Error Bounds for Variational Inequalitiesread_more
HG G 19.1
29. Oktober 2012
16:30-18:00 		Prof. Dr. Cordian Riener
Universität Konstanz
Details

Optimization Seminar

Titel Bounding hard problems with semi-definite programming using symmetries
Referent:in, Affiliation Prof. Dr. Cordian Riener, Universität Konstanz
Datum, Zeit 29. Oktober 2012, 16:30-18:00
Ort HG G 19.1
Abstract Within recent years, a lot of nice results have been obtained by applying semidefinite relaxations to hard problems. An essential step in actually solving these relaxations is to use further knowledge about additional properties of the problem such as symmetries. The talk will introduce the ideas of how to exploit symmetries in problems which can be expressed as polynomial optimization problems.
Bounding hard problems with semi-definite programming using symmetriesread_more
HG G 19.1
5. November 2012
16:30-17:30 		Prof. Dr. Jeffrey T. Linderoth
University of Wisconsin-Madison, USA
Details

Optimization Seminar

Titel Mixed-Integer Nonlinear Programs with On-Off Constraints
Referent:in, Affiliation Prof. Dr. Jeffrey T. Linderoth, University of Wisconsin-Madison, USA
Datum, Zeit 5. November 2012, 16:30-17:30
Ort HG G 19.1
Abstract We study optimization problems that have both nonlinear functional relationships between decision variables and 0-1 indicator variables that turn on and off these relationships. Problems of this class occur in many areas, including statistics, financial engineering, and engineering design. After reviewing earlier work on a reformulation technique applicable to the case when the nonlinear functions are separable, we discuss on-going research aimed at attacking the non-separable case. Our primary focus will be on the case when the nonlinearities are quadratic.
Mixed-Integer Nonlinear Programs with On-Off Constraintsread_more
HG G 19.1
12. November 2012
16:30-18:00 		Prof. Dr. Andreas Krause
Computer Science of ETH Zurich
Details

Optimization Seminar

Titel Sequential Decision Making in Uncertain Domains via Adaptive Submodularity
Referent:in, Affiliation Prof. Dr. Andreas Krause, Computer Science of ETH Zurich
Datum, Zeit 12. November 2012, 16:30-18:00
Ort HG G 19.1
Abstract Solving sequential decision problems under partial observability is a fundamental but notoriously difficult challenge. In this talk, I will introduce the new concept of adaptive submodularity, generalizing the classical notion of submodular set functions to adaptive policies. We prove that, if a problem satisfies this property, a simple adaptive greedy algorithm is guaranteed to be competitive with the optimal policy. The concept allows to recover, generalize, and extend existing results in diverse applications, including sensor management, viral marketing, and active learning. I will focus on two case studies. In an application to Bayesian experimental design, we show how greedy optimization of a novel adaptive submodular criterion outperforms standard myopic techniques based on information gain and value of information. I will also discuss how adaptive submodularity can help to address problems in computational sustainability by presenting results on conservation planning for three rare species in the Pacific Northwest of the United States.
Sequential Decision Making in Uncertain Domains via Adaptive Submodularityread_more
HG G 19.1
19. November 2012
16:30-18:00 		Prof. Dr. Mark H.A. Davis
Imperial College London
Details

Optimization Seminar

Titel Risk Management with Behavioural Factors
Referent:in, Affiliation Prof. Dr. Mark H.A. Davis, Imperial College London
Datum, Zeit 19. November 2012, 16:30-18:00
Ort HG G 19.1
Abstract In several countries a major factor contributing to the current economic crisis was massive borrowing to fund investment projects on the basis of, in retrospect, grossly optimistic valuations. We develop an approach to project valuation and risk management in which 'behavioural' factors can be explicitly included. An appropriate framework is risk-neutral valuation based on the use of the numeraire portfolio -- the 'benchmark' approach advocated by Platen and Heath (2006). We start by discussing the ingredients of the problem: 'animal spirits', financial instability, market-consistent valuation, the numeraire portfolio and structural models of credit risk. We then study a project finance problem in which a bank lends money to an entrepreneur, collateralized by the value of the latter's investment project. This contains all the components of our approach in a simple setting and illustrates what steps are required. In a final section, we outline the computational and econometric problems that arise in real applications. This is joint work with Sebastien Lleo and Grzegorz Andruszkiewicz [http://ssrn.com/abstract=2038327]
Risk Management with Behavioural Factorsread_more
HG G 19.1
* 26. November 2012
17:15-18:00
Details

Optimization Seminar

Titel No Seminar in favor of the Introductory Lecture of Mustafa H. Khammash: Mathematical Modelling, Analysis, and Computer Control of Gene Expression
Referent:in, Affiliation
Datum, Zeit 26. November 2012, 17:15-18:00
Ort HG F 30
No Seminar in favor of the Introductory Lecture of Mustafa H. Khammash: Mathematical Modelling, Analysis, and Computer Control of Gene Expression
HG F 30
3. Dezember 2012
16:30-18:00 		Prof. Dr. Alexander Martin
Friedrich-Alexander-Universität Erlangen-Nürnberg
Details

Optimization Seminar

Titel May Mixed Integer Programming help to sove MINLPs? A Case Study from Gas Network Optimization
Referent:in, Affiliation Prof. Dr. Alexander Martin, Friedrich-Alexander-Universität Erlangen-Nürnberg
Datum, Zeit 3. Dezember 2012, 16:30-18:00
Ort HG G 19.1
Abstract Mixed integer nonlinear programs are currently in the focus of mathematical optimization. The sizes of problems that are actually solvable, however, are still low compared to those solvable when only linear inequalities are involved. We propose a discretization technique to approximate the nonlinear functions by linear ones in an iterative way to meet some given tolerance. This way we are able to apply MIP solution techniques. We will demonstrate the success of this approach on a particular class of applications resulting from gas network optimization.
May Mixed Integer Programming help to sove MINLPs? A Case Study from Gas Network Optimizationread_more
HG G 19.1

Hinweise: mit einem Stern gekennzeichnete Ereignisse (*) zeigen an, dass die Zeit und/oder der Ort von der üblichen Zeit und/oder dem üblichen Ort abweichen und wenn Sie möchten, können Sie den iCal/ics-Kalender abonnieren.

Organisatoren:innen: Komei Fukuda, Bernd Gärtner, Diethard Klatte, John Lygeros, Manfred Morari, Karl Schmedders, Robert Weismantel

Archiv: FS 22 FS 20 HS 19 FS 19 HS 18 FS 18 HS 17 FS 17 HS 16 FS 16 HS 15 FS 15 HS 14 FS 14 HS 13 FS 13 HS 12 FS 12 HS 11 FS 11 HS 10 FS 10 HS 09

JavaScript wurde auf Ihrem Browser deaktiviert