Durham Training School

The 2018 training school for UTOPIAE will take place 2–6 July 2018 at Durham University. At registration it will be possible to book accommodation at Collingwood College, with the training school itself taking place in Earth Sciences (ES231, see map below).

To register, please click here (deadline: 15th June)

Getting here

Details about how to get to Durham are available from the university website here.

Note: the free bus service from Durham Tees Valley airport has been dropped in the past. For those travelling by plane:

According to the Durham ESRs’ experience, Newcastle is the best choice.

From Durham train station it is about 5 min walk to Durham Bus station, from where you can take bus no. 6 which stops at “South Road Colleges”, near the accommodation site (see the map below).

Key locations

Click the markers in the Google Map below for details about key locations for the workshop:

Provisional timetable

Note that all attendees are welcome to attend the key skills training sessions. The only private UTOPIAE sessions are 2pm–5pm on Tuesday and 1pm–4pm on Wednesday. On Tuesday attendees who are not part of UTOPIAE are welcome to present short talks on their work in a parallel session: if you would like to do this, please tick the option at registration and we will contact you for a talk title.

To register, please click here (deadline: 15th June)

There will be tea/coffee breaks throughout the day (not shown).

Monday 2nd  
9.00 - 11.00 ‘Introduction to Bayesian statistics’
    Georgios Karagiannis, Durham University
11.00 - 13.00 ‘Introduction to the theory of imprecise probability’
    Erik Quaeghebeur, TU Delft
13.00 - 13.40 Buffet lunch
13.40 - 15.40 ‘Statistical modelling with imprecise probabilities’
    Thomas Augustin, University of Munich
16:15 - Tour of Durham Castle
   
   
Tuesday 3rd  
   
9.00 - 11.00 ‘Markov chains with imprecise probabilities’
    Gert de Cooman, Ghent University
11.00 - 13.00 ‘Introduction to reliability theory’
    Lisa Jackson, Loughborough University and Frank Coolen, Durham University
13.00 - 14.00 Buffet lunch
14.00 - 17.00 Afternoon ESR activity (UTOPIAE members only)
  Non-UTOPIAE participant short talks
   
   
Wednesday 4th  
   
9.00 - 11.00 ‘Sampling from complex probability distributions’
    Louis Aslett, Durham University
11.00 - 12.00 ‘Key skills training: Reviewing papers’
    Jochen Einbeck, Durham University
12.00 - 13.00 Buffet lunch
13.00 - 16.00 Project working group presentations
16.00 - 17.00 ‘Public lecture: Randomness’
    Gert de Cooman, Ghent University
   
   
Thursday 5th  
   
9.00 - 11.00 ‘Engineering pathways for quantifying inconsistent information’
    Michael Beer, Leibniz Universität Hannover
11.00 - 13.00 ‘Simulation methods for the analysis of complex systems’
    Edoardo Patelli, University of Liverpool
13.00 - 14.00 Buffet lunch
14.00 - 17.00 Key skills training (Michael Beer, Leibniz Universität Hannover)
       1. ‘Publish successfully’
       2. ‘Careers advice’
  (UTOPIAE supervisors: Network meeting)
18.30 - Dinner @ Hatfield College
   
   
Friday 6th  
   
9.00 - 11.00 ‘Reduction approaches for the Bayesian inference of model parameter: methods and examples’
    Olivier Le Maître, CNRS, and Pietro Congedo, INRIA
11.00 - 13.00 ‘Analysis and model validation starting with experimental measurements’
    Olivier Chazot and Thierry Magin, von Karman Institute
To register, please click here (deadline: 15th June)

Talks

Monday, 9am – 11am

Introduction to Bayesian statistics

Georgios Karagiannis, Durham University

We introduce basic concepts of Bayesian theory. We introduce the Bayesian paradigm, and discuss the subjective interpretation of probability. We study methods to incorporate prior information in the Bayesian statistical analysis. From a decision theory perspective, we study parametric and predictive inference, and particularly the design of point estimators, credible sets, and hypothesis tests. Finally, we discuss how the these concepts can be applied to build complex Bayesian models able to address realistic problems.


Monday, 11am – 1pm

Introduction to the theory of imprecise probability

Erik Quaeghebeur, TU Delft

The theory of imprecise probability is a generalization of classical ‘precise’ probability theory that allows modeling imprecision and indecision. This is a practical advantage in situations where a unique precise uncertainty model cannot be justified. This arises, for example, when there is a relatively small amount of data available to learn the uncertainty model or when the model’s structure cannot be defined uniquely. The tools the theory provides make it possible to draw conclusions and make decisions that correctly reflect the limited information or knowledge available for the uncertainty modeling task. This extra expressivity however often implies a higher computational burden.

The goal of this lecture is to primarily give you the necessary knowledge to be able to read literature that makes use of the theory of imprecise probability. A secondary goal is to provide the insight needed to use imprecise probabilities in your own research. To achieve the goals, we will present the essential concepts and techniques from the theory, as well as give a less in-depth overview of the various specific uncertainty models used. Throughout, examples will be used to make things concrete. We build on the assumed basic knowledge of classical probability theory, for which parts of the “Introduction to Bayesian statistics” lecture provides a refresher, and prepare the way for many of the subsequent lectures of the School.


Monday, 1:40pm – 3:40pm

Statistical modelling with imprecise probabilities

Thomas Augustin, University of Munich

The presentation gives a first introduction into statistical modelling with imprecision.

The first part considers some typical applications in the context of model imprecision. We first show how imprecise probabilities provide a natural superstructure upon (frequentist) robust statistics. Then we look at the use of imprecise priors in Generalized Bayesian inference, where we discuss the proper handling of near prior ignorance as well as of prior-data conflict.

The second part focuses on data imprecision, i.e. situations where data are not observed in the resolution originally intended. Following recent work by Couso \& Dubois, we distinguish between ontic and epistemic data imprecision (precise observation of something genuinely imprecise versus imprecise observation of something genuinely precise) and discuss for both situations first approaches for modelling coarse and missing data with imprecise probabilities.


Tuesday, 9am – 11am

Markov chains with imprecise probabilities

Gert de Cooman, Ghent University

Abstract to follow.


Tuesday, 11am – 1pm

Introduction to reliability theory

Lisa Jackson, Loughborough University and Frank Coolen, Durham University

In this lecture we introduce basic concepts of reliability theory, with focus on system reliability. We discuss and illustrate reliability quantification for a range of scenarios, including dependent failures, phased missions and competing risks. We discuss statistical methods for reliability inferences, including imprecise probability approaches. Finally, we will briefly discuss several challenges for research, related to resilience, lack of information and upscaling of theory and methods to large real world applications.


Wednesday, 9am – 11am

Sampling from complex probability distributions

Louis Aslett, Durham University

The need to sample from complex probability distributions arises in a multitude of settings, from simulation of an engineering system through to Bayesian posterior inference. In this lecture we will introduce standard Monte Carlo methods and continue through to modern Markov Chain Monte Carlo (MCMC) techniques, carefully building up a little introductory level theory, several modern methodologies and practical implementation. The context will be kept as general as possible, focusing on an arbitrary target distribution of interest, enabling the techniques learned to be applied in any setting where a probability distribution can be specified up to a constant of proportionality.


Wednesday, 11am – noon

Key skills training: Reviewing papers

Jochen Einbeck, Durham University

This key skills session discusses several aspects of scholarly peer review, with focus on the referee’s role in this central element of scientific publishing. The main part of this session will be arranged in form of a workshop where participants identify examples of good and bad practice from a set of case studies.


Wednesday, 4pm – 5pm

Public lecture: Randomness and imprecision

Gert de Cooman, Ghent University

The randomness of a sequence of numbers can be defined in many ways. The talk begins with a short survey of the most common definitions of randomness and their relationships, and then focuses on a powerful and intuitive martingale-theoretic definition first suggested by Ville, and further refined by Schnorr and Levin. It essentially requires that there should be no (in some way computationally achievable) strategy for gambling on the successive outcomes in the sequence that allows a player, Skeptic, to become infinitely rich without borrowing. Interestingly, this betting approach allows for a generalisation towards interval (or imprecise) probabilities. As is often the case with the mathematics of imprecise probabilities, this allows for new ideas and structures to emerge, and takes us to a new vantage point from where it becomes easier to appreciate the subtleties and intricacies associated with the precise limit case where intervals reduce to numbers.

Bio: Gert de Cooman is Professor of Uncertainty Modelling and Systems Science in Ghent University, Belgium. He was a Grey College Alan Richards Fellow in Mathematics (2006) and Visiting Professor at Durham University from 2014 till 2017. His main research interest lies in dealing with robustness, imprecision and indecision in probabilistic modelling, and in particular the mathematical study of lower expectation functionals, the treatment of imprecision in stochastic processes using supermartingales, the foundations of decision making using sets of desirable gambles and choice functions, and most recently the study of the relation between randomness and imprecision. He is a founding member and former President of SIPTA (the international Society for Imprecise Probabilities: Theories and Applications), and is currently a member of its Executive Committee.


Thursday, 9am – 11am

Engineering pathways for quantifying inconsistent information

Michael Beer, Leibniz Universität Hannover

In engineering practice data and information are normally not available in an ideal form for mathematical quantification with with probabilistic models. In those cases imprecise probabilities offer a framework to remedy the problem. However, depending on the specific problem, there is still a large range of options on how to put up the model. In some cases, we may have strong evidence for the type of distribution, and vagueness exists only on the probabilistic model parameter values. In other cases, we may have conflicting expert assessments on distribution types and some data of only poor quality. Occasionally, we may only have a rough idea about the range of structural or environmental parameters associated with a credibility expressed as subjective probability. In each particular case the analyst needs to decide how to establish the model using the rich framework of imprecise probabilities, matching the technical context of the problem at hand. In this presentation we will discuss practical engineering pathways for solving this problem.


Thursday, 11am – 1pm

Simulation methods for the analysis of complex systems

Edoardo Patelli, University of Liverpool

Analysis of systems is traditionally performed adopting different tools reliability block diagrams, fault tree and success tree methods, failure mode and effect analysis, etc… These traditional approaches are difficult or impossible to apply to large complex systems due to the complex and tedious calculations for finding minimal path sets and cut sets. Instead, simulation approaches can be used to investigate large and complex systems and for obtaining numerical solutions where analytical solutions are not available. In particular, simulation methods allow to consider explicitly the effect of uncertainty and imprecision on the system under investigation providing a powerful tool for risk analysis which allows better decision making under uncertainty.

A simple and generally applicable simulation approach, enhanced for multi-state systems of any topology is presented. Each component is defined as a Semi-Markov stochastic process and via discrete-event simulation. The principles of flow conservation are invoked to determine flow across the system for every performance level change of its components using the interior-point algorithm. This eliminates the need for cut-set definition and overcomes the limitations of existing techniques. The methodology can also be exploited to account for effects of transmission efficiency and loading restrictions of components on system reliability and performance. The approach is simple and generally applicable to systems, including those with limited maintenance teams, reconfiguration requirements, and multiple commodity flows

In addition, efficient simulation approaches will be presented to estimate the reliability of systems based on survival signature. The simulation approaches are generally applicable to any system configuration and allow to consider different representation of the uncertainties (i.e. distribution, intervals, fuzzy etc) and system multi-state components (i.e. repairable components).


Thursday, 2pm – 5pm

Publish successfully

Michael Beer, Leibniz Universität Hannover

The presentation will highlight some key issues to take of when preparing journal papers. Advice will be provided on how to select the right journal for the paper, how to define the right paper type, and how to work out a research paper with emphasis on the key items that make it a research paper. In addition to these basic items, information will be provided on how to address reviewer comments, how to understand editorial decisions, and how to deal with conflicting information and problems. In overall, the information provided targets at increasing the chance for success when submitting journal papers. This presentation is based on own experience as an author, reviewer and handling and associate editor for several international journals.

Followed by:

Careers advice

Michael Beer, Leibniz Universität Hannover

Going for an academic career means to set out for a dynamic pathway with often unexpected obstacles and restrictions and also chances. Getting prepared early helps to reduce uncertainties and to increase chances significantly. The presentation will highlight a number of facts that are crucial for success but are often not considered or considered too late. It will be explained how to develop an impactful CV as a basis for applying for academic jobs worldwide. It will also be explained how applications are assessed, and how to prepare them to make the most imporatnt items clear to the panels and referees. This presentation is based on own experience in different academic systems and on experience from assessing applications and writing and assessing reference letters.


Friday, 9am – 11am

Reduction approaches for the Bayesian inference of model parameter: methods and examples

Olivier Le Maître, CNRS, and Pietro Congedo, INRIA

Abstract to follow.


Friday, 11am – 1pm

Analysis and model validation starting with experimental measurements

Olivier Chazot and Thierry Magin, von Karman Institute

The lecture presents, in a first part the conception of experiments for the investigations of aerothermodynamic phenomena in high enthalpy facilities and plasma wind tunnels. It develops the framework in which an experiment is relevant for representing a real flight situation and underlines its limitations.

The second part focus on the experimental measurements with their techniques and their related data processing tools. It exposes the databases that could be generated with the definition of the corresponding basic uncertainty margins.