Operational Research and Statistics Seminars 2014-2015
All seminars will commence at 12:10pm in room M/2.06, The Mathematics Building, Cardiff University, Senghennydd Road (unless otherwise stated).
1 October 2014
Speaker: Dr Trivikram Dokka (Lancaster)
Title: New polyhedral results for the three-index assignment problem.
Abstract: PDF Download
20 October 2014 at 17:30 in Room M/0.40
Speaker: Prof. Michael Carter (Toronto).
Title: Health System Patient Flow Simulation Model.
Abstract: Our cross-sector patient flow model is a system dynamics simulation focused on the flow rates of patients between health system sectors, and the feedback structures around them. It takes a whole-system, strategic perspective, and is designed to produce output that captures the direction and magnitude resulting from policy changes around patient pathways and service levels.
The model was developed within a geographically large health region of around one million people. It was constructed based on consultations with four expert panels: health policy leaders, acute, institutional and home and community care. The first panel identified the policy levers targeting flows of patients between healthcare sectors, while the other three panels were engaged to develop causal loop diagrams (CSD) that explained admission and discharge flows of their respective sectors. Qualitative input from these panels was merged with health system data to develop a stock-and-flow structure of the health region. Patients are grouped into cohorts by age, sex, clinical condition, referral source and discharge destination. The model output was validated against admission and discharge data collected over a four year period.
The model has been applied to both qualitative (CSD) and quantitative (stock and flow) decision problems. The CSD model was used by the Ontario Ministry of Health and Long Term Care to test a new slow stream rehab option for complex patients. The model demonstrated several negative unintended consequences and led the policy team to revise their recommendations. The quantitative model has been used to validate the Ontario Stroke Strategy to estimate length of stay improvements. We are currently working with the Ministry to make the tool more widely accessible.
5 November 2014
Speaker: Dr Ben Torsney (Glasgow),.
Title: Optimal Design, Lagrangian and Linear Model Theories: Further Developments on a Fusion.
20 November 2014
Speaker: Dr Vadim Lozin (Warwick).
Please note change of room for this - M/2.01.
Title: Combinatorics and algorithms for augmenting graphs.
Abstract: The notion of augmenting graphs generalizes the Berge's idea of augmenting chains that has been used by Edmonds in his celebrated solution of the maximum matching problem. This problem is a special case of the more general problem of finding a maximum independent set in a graph. Recently, the augmenting graph approach has been successfully applied to solve the maximum independent set problem in various other special cases. However, our knowledge of augmenting graphs is still very limited, and we do not even know what the minimal *infinite* classes of augmenting graphs are. In this talk, we give an answer to this question and apply it to extend the area of polynomial-time solvability of the maximum independent set problem.
28 January 2015
This will take place in room M/2.06.
Speaker: Prof. Robert John (Nottingham).
Title: Type-2 Fuzzy Logic in Decision Support
Abstract: This talk will provide an overview of Bob's research in type-2 fuzzy logic and its application in Decision Support. Type-2 fuzzy sets are fuzzy-fuzzy sets - that is, where the fuzzy set has membership grades that are themselves fuzzy sets, rather than numbers in [0,1]. Fuzzy sets (type-1) have had significant success in control applications but by their very definition are not particularly 'fuzzy' and struggle in applications that attempt to mimic human reasoning in decision support systems. Introduced in 1975, type-2 fuzzy logic really started to grow in the late '90s led by Bob and Jerry Mendel. In the intervening period the number of type-2 papers and researchers has grown considerably. This talk will introduce the audience to type-2 fuzzy logic and provide a brief history.
Bob will describe practical application of his work in decision support, such as the aggregation of uncertain information, supply chain modelling and medical diagnosis.
Bob John has a BSc in Mathematics, a MSc in Statistics and a PhD in Fuzzy Logic. He worked in industry for 10 years as a mathematician and knowledge engineer developing knowledge based systems for British Gas and the financial services industry. Bob spent 24 years at De Montfort University in various roles including Head of Department, Head of School and Deputy Dean. He led the Centre for Computational Intelligence research group from 2001 until 2012. Bob joined Nottingham in 2013 where he led on the LANCS initiative and Heads up the research group Automated Scheduling, Optimisation and Planning (ASAP) in the School of Computer Science. The LANCS Initiative is built on a collaboration between four U.K. Universities: Lancaster, Nottingham, Cardiff and Southampton. The research group carries out multi-disciplinary research into mathematical models and algorithms for a variety of real world optimisation problems. It has 8 academic staff, 9 researchers and over 30 PhD students.
11 February 2015
Speaker: Dr Fabricio Oliveira (Rio de Janeiro).
Title: Optimising under uncertainty: an introduction and applications in healthcare related problems.
Abstract: In this talk we will discuss the importance, as well as some of the available tools, to consider the stochastic nature of input parameters in optimisation problems. It is well known that the static nature of optimisation problems makes it difficult to support decision making when the input data is subject to uncertainty. We will present how one can incorporate such uncertainty by means of stochastic and robust optimisation, using examples of current applications in healthcare related problems. Some previous background in optimisation is of good value, but not necessarily mandatory. Hopefully, at the end of the talk, the audience will be able to understand how it is possible to include uncertainty in optimization-based decision support tools to improve the decision-making process.
18 February 2015
Speaker: Christian Henning (UCL).
4 March 2015
Speaker: Dr Dmitrii Pasechnik (Oxford).
18 March 2015 at 15:00
Speaker: Haeran Cho (Bristol).
15 April 2015
Speaker: Dr Vladimir Deineko (Warwick).
Title: Special structures in polynomially solvable cases: Is there much in common?
Abstract: In our talk we present a survey of polynomially solvable cases of NP-hard problems with an emphasis on common structures in these cases. We concentrate on the cases where specially structured matrices are involved. It most considered cases permuting rows and columns of specially structured matrices destroy the properties needed. We discuss the arising recognition problems and pose quite a few open questions from this exciting area of research.
29 April 2015
Speaker: Prof. Mark Kelbert (Moscow).
Title: Shannon's entropy power inequality and weighted differential entropies.
Abstract: We establish a number of new inequalities for weighted differential entropies and analyze in details a Bayesian problem of estimating probability of success in a series of trials with binary outcomes. In particular, the weighted Rao-Cram\'er inequality is presented.
6 May 2015
Speaker: Prof. Mark Girolami (Warwick).
27 May 2015
Speaker: Dr Timm Oertel (Zurich).
Title: A polyhedral Frobenius theorem with applications to integer optimization.
Abstract: We prove a representation theorem of projections of sets of integer points by an integer matrix W. Our result can be seen as a polyhedral analogue of several classical and recent results related to the Frobenius problem. Our result is motivated by a large class of non-linear integer optimization problems in variable dimension. Concretely, we aim to optimize f(Wx) over the set of integers in P, where f is a non-linear function, P is a n-dimensional polyhedron and W is a d x n matrix. As a consequence of our representation theorem, we obtain a general efficient transformation from the latter class of problems to integer linear programming. Our bounds depends polynomially on various important parameters of the input data leading, among others, to first polynomial time algorithms for several classes of non-linear optimization problems.