The following seminars will be held in the academic year 2008/9.

Thursday 28 May 2009 - 3pm to 4.30pm

On Non-cooperative Game Theory

Speaker: Professor Gábor Kassay, Babes-Bolyai University Cluj, Romania

Room: 886, Graham Hills Building.

We introduce the concept of two-player non-cooperative game and pay attention to the special case of zero-sum games (i.e., the gain of each player equals the loss of his opponent) where the set of pure strategies (also called actions) for both players are finite sets. The latter is called a matrix game.

Examples for matrix games will be provided in order to show their use in real life problems.

Then, we introduce the sets of mixed strategies for both players and define the so-called optimal strategies. The first result stating that every matrix game admits optimal strategies for both players has been obtained by J. von Neumann in 1928. It follows from his result that every matrix game is solvable. We extend our investigations for not necessarily matrix games (i.e., the set of mixed strategies for any player can be infinite) and give a short overview of the most famous results on the existence of optimal strategies for these games.

We conclude our presentation with the concept of n-person game, in which framework Nash equilibrium points and Nash's theorem play a crucial role.

Tuesday 26 May 2009 - 3pm to 4.30pm

Towards the Development of a Comprehensive Vessel Traffic Risk Management Procedure

Speaker: Professor J. René van Dorp, Department of Engineering Management and Systems Engineering, The George Washington University

Room: 886, Graham Hills Building.

Abstract: Is it safer for a New Orleans river gambling boats to be underway than to be dockside? Does oil transportation risk reduce by lowering wind restrictions from 45 to 35 knots at Hinchinbrook Entrance for laden oil tankers departing Valdez, Alaska? Should the International Safety Management (ISM) code be implemented fleet wide for the Washington State Ferries in Seattle or does it make more sense to invest in additional life craft? Can ferry service be expanded in San Francisco Bay alleviate in a safe manner to relieve high? These risk management questions were raised in a series of projects spanning a time frame of more than 10 years. They were addressed using a risk management analysis methodology developed by a consortium of universities. In this talk we shall review this methodology which integrates simulation of Maritime Transportation Systems (MTS) with incident/accident data collection, expert judgment elicitation and a consequence model. We shall describe recent advances that were made to the methodology in the context of a two-year oil transportation risk study conducted from 2006-2008 in the Puget Sound and surrounding waters. An application of this methodology shall be presented comparing the risk reduction effectiveness analysis of a one-way zone requirement, escorting requirements and a double hull requirement.

Friday 22 May 2009 - 1pm to 2pm

Learning additive value functions: The Sound-Score method

Speaker: Professor József Dombi, University of Szeged, Department of Algorithms and Artificial Intelligence

Room: 886, Graham Hills Building.

Multicriteria decision analysis studies decision problems in which the alternatives are evaluated on several dimensions on view points. We presented a method called Sound-Score system for ranking finite set of alternatives evaluated on multiple criteria. The method builds a set of additive value functions compatible with preference information composed of a partial preorder. The method generalizes both UTA, UTAGMS, GRIP method. The Sound-Score method can be composed to the AHP method which requires from the decision maker preference information on p… comparison. The Sound-Score method is not restricted to monotone increasing (or decreasing) value function, if we have information of the maximum points. The advantage of the method is that the preference information does not need to be complex. Using the method we can also incorporate the knowledge on the intensity of the preference, as the knowledge on the preference of the weights. By the end of the lecture we show a demonstration program.

Monday 30 March 2009 - 4pm to 5pm

Supporting Project Prioritisation in a Multi-Organisational Setting: insights from a facilitated modelling intervention

Speaker: Professor Alberto Franco, Warwick Business School, University of Warwick, Coventry

Room: 886, Graham Hills Building.

This talk will discuss the practical experience gained from the design and use of a facilitated modelling intervention in an English borough. The purpose of the intervention was to provide support to a multi-organizational group tasked with tackling problems associated with teenage pregnancy. The intervention adopted a multi-methodology approach by combining facilitated, computer-supported causal mapping and multi-criteria decision analysis to develop and prioritise a number of projects aimed at alleviating teenage pregnancy issues. The talk will describe the process of developing and applying the intervention approach, and provide an evaluation of its perceived impact by the client group. Overall, the experience suggests some issues concerning the joint deployment of facilitated problem structuring and decision analysis in multi-organisational settings. Directions for further research are also identified.

Monday 23 March 2009 - 5pm to 6pm

Assessing Tools which involve Explicit or Implicit Questioning in Facilitated Decision Modelling:
comparing SSM with an exercise based upon a board game (RISK)

Speaker: Dr. Giles Hindle, Warwick Business School, The University of Warwick

In the practice of facilitated decision modelling, we often wish to generate ideas about the nature of an organisation or business unit and how a group of participants might want the unit to look in the future. In SSM, this will usually involve conceptualising the organisation or business unit as a Human Activity System and developing Root Definitions and Conceptual Models (both descriptive and creative models). In a recent SSM study for the National Public Health Service of Wales, the author intended to do just that in three facilitated workshops. However, due to perceived difficulties with preparatory modelling, the author, in partnership with the Director of Operations, decided to ditch the conceptual language of SSM and develop an exercise based upon the board game RISK. We feel this provided a much better workshop experience for the participants, whist still generating the qualitative data we needed on the nature of the “system” and alternative system designs for the future. The main difference between the exercise and an SSM-approach is that the game raised fundamental questions implicitly, rather than explicitly.

Monday 16 March 2009 - 5.30pm to 6.30pm

Evaluation of group model building

Speaker: Etienne Rouwette, Radboud University, Nijmegen, The Netherlands

Group model building is the construction of system dynamics models in direct participation with decision makers and experts. In the Netherlands group model building has been applied to, among others, problems in criminal justice, health care, defence, taxes, and telecommunication. I will show a general outline for a group model building process and illustrate this with examples from several cases, using both qualitative and quantitative applications. Evaluation of these cases shows a number of surprising insights, for example a discrepancy between self-assessment of learning and evaluations on the basis of pretest - posttest measurements. We then discuss theories from cognitive and social psychology that might explain observed results of group model building.

Short biography of speaker: Etiënne Rouwette is associate professor at the Methodology group of the Faculty of Management Sciences, Radboud University Nijmegen (the Netherlands). He is the author of several articles on evaluation of group decision support, including group model building, simulation gaming and electronic meeting support. Since 1993 he is involved in consultancy applications using group model building and other approaches to group decision support.

Thursday 12 March 2009 - 5.15pm to 6.30pm

Computational results of an O*(n4) volume algorithm

Speaker: Professor István Deák, Corvinus University, Budapest, Hungary

Room: 886, Graham Hills Building.

Computing the volume of an n-dimensional body is NP-hard, but computing an estimate of it can be done in polynomial time. An O*(n4) randomized volume algorithm has been presented for convex bodies by Lovász and Vempala (2004).  Essentially the algorithm is a series of Monte Carlo integrations, so this is a non-deterministic algorithm. In this lecture we describe a computer implementation of the volume algorithm, and present some computational results for convex bodies in dimensions n=2, 3, &, 10.

Presently the main application of this algorithm seems to be stochastic optimization: a randomized algorithm can be produced when optimizing a linear objective function over a convex body. In statistics, data mining and for limited databases probability densities may be defined and visualized after computing volumes of convex bodies spanned by data, quantification of risk factors become possible. An important part of the volume algorithm is generating random points uniformly in a convex body; this was practically impossible before and is useful in all kinds of n-dimensional simulation. 

Thursday 26 February 2009 - 4 to 5pm

Pessimism and risk aversion under rank dependent utilities

Speaker: Professor Isaac Meilijson, School of Mathematical Sciences, Tel Aviv University, Tel Aviv, Israel

Room: 859F, Graham Hills Building.

Von-Neumann & Morgenstern and Savage axioms lead to the existence for each of us of a utility function U such that we prefer a lottery X to another Y if and only if E[U(X)]>=E[U(Y)]. There is evidence that Homo sapiens don't adhere tightly to these axioms. For example (Allais, Kahnemann-Tversky), if £1M is preferred to letting a fair coin decide between zero and £5M, then necessarily a lotto ticket with prize £1M with probability 1/1M (else zero) must be preferred to £5M with probability 1/2M.

Experiments with human decision makers show conflict. People don't really differentiate between 1/1M and 1/2M but are attracted to £5M versus £1M. However, probabilities 1/1 or 1/2 are very seriously different.

Under the rank-dependent utility model agents are characterised by a pair of functions, the same utility U as in Von-Neumann & Morgenstern that distorts money, added to a probability perception function that distorts probabilities.

The talk deals with pessimists and tries to characterise risk averse agents and attitudes to risk taking. Technical emphasis will be placed on stochastic orders modelling risk and various types of risk aversion, not philosophical emphasis on axioms leading to RDU.

Friday 20 February 2009 - 3.30pm

Operational Research in Consultancy

Speaker: Finlay Buchanan of Cap Gemini

Room: 886, Graham Hills Building.

Finlay will describe life and work in the UK's largest consulting operational research group. He will let you into the key on-going secrets from the A-Z of consulting buzzwords and tell you about the best-ever operational research project in the entire world!

Monday 9 February 2009 - 4.15pm to 5.30pm

Optimization of Electricity Portfolios

Speaker: Professor István Maros, Imperial College, London, UK and University of Pannonia, Veszprém, Hungary

Room: 859F, Graham Hills Building.

Regional electricity distribution companies (DCs) have the legal obligation to provide electricity for their customers in every half hour period of the year. Companies purchase electricity from suppliers (domestic power stations, import) based on yearly contracts. Suppliers make several different contract offers. A contract offer contains time periods covered and the level of service (measured in some units). DCs purpose is to choose a combination of contract offers that minimizes costs and satisfies demands as close as possible. The talk presents three models that lead to large scale optimization problems. Some interesting mathematical properties of the models make it possible to solve them very efficiently. The technique, which will also be described, enables the user to evaluate several versions of any of the models as solution times can be reduced drastically.

Thursday 20 November 2008 - 3pm to 5pm

Stochastic geometry and telecommunications modelling

Speaker: Dr. Sergei Zuyev, Reader, University of Strathclyde, Department of Statistics and Modelling Science

Room: 886, Graham Hills Building.

Dr. Zuyev will give his talk in two parts. First part will focus more on application, while the second part more on mathematical issues of this application. There will be break between the two parts of the talk. Second part will start at 16:00.

ABSTRACT: Stochastic geometry gradually becomes a necessary theoretical tool to model and analyse modern telecommunication systems, very much the same way the queuing theory revolutionised studying the circuit switched telephony in the last century. The reason for this is that the spatial structure of most contemporary networks plays crucial role in their functioning and thus it has to be properly accounted for when doing their performance evaluation, optimisation or deciding the best evolution scenarios.  The talk will present some stochastic geometry models and tools currently used in studying modern telecommunications.  We outline specifics of wired, wireless fixed and ad-hoc systems and show how the stochastic geometry modelling helps in their analysis and optimisation.

Measures everywhere!
 
Branches of science using measures are abounding: length, volume, mass, electric charge, probability, point processes are all examples of measures. Looking at the set of measures as a convex cone in a Banach space we derive specific first and second order necessary conditions for extrema of functionals of measures. We discuss applications in numeric approximation, probability, statistics and optimal experimental design. The obtained explicit formula for the gradient allows us to develop new steepest descent algorithms efficiency of which will be shown on numeric examples.
 

Thursday 9 October 2008 - 3pm

On semidefinite programming relaxations of the traveling salesman problem

Speaker: Professor Etienne de Klerk, Tilburg University, The Netherlands

Room: 886, Graham Hills Building.

Joint work with: Dmitrii V. Pasechnik, Renata Sotirov

ABSTRACT: We consider a new semidefinite programming (SDP) relaxation of the symmetric traveling  salesman problem (TSP), obtained via an SDP relaxation of the more general quadratic  assignment problem (QAP). We show that the new relaxation dominates the one in the  paper: [D. Cvetkovic, M. Cangalovic and V. Kovacevic-Vujcic. Semidefinite Programming  Methods for the Symmetric Traveling Salesman Problem. In Proceedings of the 7th International  IPCO Conference on Integer Programming and Combinatorial Optimization, 1999, 126136, Springer-Verlag, London, UK.] Unlike the relaxation of Cvetkovic et al., the new SDP relaxation is not dominated by the linear programming relaxation with sub-tour elimination constraints (the Held-Karp bound).

Note to the audience: the first part of the talk will be a general introduction to the theory and applications of the traveling salesman and quadratic assignment problems. The applications include facility location planning, routing problems, and circuit board design.