Instructions for presenters:see instructions page.
Program at a glanceThe registration desk is open on every morning and early afternoon of the main conference days. In addition, the desk is open on Sunday from noon until 6:00 pm.
Tutorial programThe ECSQARU is preceded by a Tutorial day on Sunday July 7 (2013).
The main ECSQARU conference program runs from Monday July 8 until Wednesday July 10 (2013).
- Each day of the conference starts with an invited talk (see speakers and abstracts)
- Global program:
- Monday July 8: Conference (all day) followed by welcome reception
- Tuesday July 9: Conference (all day) followed by dinner
- Wednesday July 10: Conference (all day). During the closing session the 'best paper award' and the 'best student paper award' will be presented.
- Detailed program; or just a list of accepted papers.
Best Paper Jury
The jury for selecting the best paper is monitored by L.C. van der Gaag and consists of:
- G. de Cooman
- A.P. Dawid
- P. Grünwald
- G. Coletti
- S. Destercke
- L. Godo
Each day of the conference starts with an invited talk. Our three invited speakers are:
- Prof. A. Philip Dawid, University of Cambridge
- Prof. Simon Parsons, Brooklyn College, New York
- Prof. Gert de Cooman, Ghent University
Conditional independence for causal reasoning (presentation)
Although conditional independence is fundamentally a probabilistic notion, it has been widely used to represent, manipulate, and suggest causal relationships. I will review the variety of ways in which this has been done, and consider the extent to which these can be given some logical justification.
The state of the argument
In the past twenty years, computational argumentation has moved from being a minority interest to a field that is increasingly seen as part of the AI mainstream. This talk will reflect on the change. It will provide an overview of the current state of the field, but one that harks back to the state of the field 20 years ago, and suggests some fruitful areas for future research.
Inference under exchangeability using sets of desirable gambles
I give a concise overview of one of the most powerful and elegant languages for representing and reasoning under uncertainty: sets of desirable gambles. I provide evidence for its simplicity, elegance and power by zooming in on a few examples: dealing with irrelevance and independence in credal networks, dealing with symmetry in general and exchangeability in particular, and studying predictive inference systems to show that the Imprecise Dirichlet Model satisfies a number of very interesting properties.
Prior to the main conference, on Sunday July 7, we have the afternoon filled with 4 tutorials. Separate registration is required for the tutorial program, and provides entrance to all 4 tutorials.
- Hans Bodlaender (Algorithmic Systems, Department of Information and Computing Sciences, Utrecht University, the Netherlands)
- Fabio Cozman (Decision making lab, Universidade de Sao Paulo, Brazil)
- Jonathan Lawry (Department of Engineering Mathematics, University Of Bristol, UK)
- Sébastien Destercke (Centre de recherche de Royallieu, Université de Technologie de Compiègne, France)
An introduction to fixed parameter tractability and kernelization (presentation)
Fixed parameter tractability is a relatively new field in algorithm research. Here, the running time of an algorithm is measured not only as a function of the size of the input but also of a second parameter of the input, which typically is significantly smaller. E.g., a problem to find a certain structure for a given input of size at most k may be NP-complete, but if k is some small fixed number, it may well be polynomial time solvable. In the talk, algorithmic and lower bound techniques and important notions from the field will be discussed. We also look at the related notion of kernelization, which intuitively amounts to preprocessing instances of a problem to equivalent instances with a guaranteed bound on the size of resulting instances.
Concepts of independence for coherent probabilities and for credal sets (presentation)
Standard probability theory, as put together by Kolmogorov, offers a venerable framework to reason with and about uncertainty. However, that theory is not the only way to deal with probability, let alone the only way to deal with uncertainty. This tutorial will focus on two alternatives to standard probability theory. The first part of the tutorial will describe the theory of coherent probabilities, where conditional probability is a primitive concept that can be used to encode a large variety of formalisms. The second part of the tutorial will describe the theory of sets of probability measures (known as credal sets), where the dogma of uniqueness of probability values is abandoned. Both theories have interesting properties and connections that are best appreciated when one investigates concepts of independence. The third part of the tutorial will focus on concepts of independence based on factorization and on irrelevance. Whenever possible, concepts of independence will be evaluated with respect to their graphoid properties and their ability to support modeling tools such as Bayesian networks and Markov random fields.
Vagueness in intelligent systems (presentation)
This tutorial will explore the potential benefits of embedding vagueness as part of formal knowledge representation in intelligent systems. By focusing on the utility of vagueness in multi-agent communications, natural language generation, consensus modelling and decision making, we will investigate different aspects of the phenomenon and outline how these are captured by a number of distinct theories. This will include, supervaluationism, many-valued logics (especially three-valued logic), and the epistemic theory of vagueness. In particular, we will consider the different roles of indeterminism (i.e. truth-gaps or borderline cases), semantic uncertainty (i.e. uncertainty about interpretations or language conventions) and typicality (i.e. similarity to prototypical cases) within an enriched concept representation framework. Bringing together these ideas we will then describe in some detail a possible representational framework incorporating several different aspects of vagueness. Along the way we will highlight relationships with other models of vagueness and uncertainty including fuzzy logic, interval fuzzy logic, random set theory, Dempster-Shafer theory and possibility theory. The potential of this approach will then be illustrated using examples of its application in artificial intelligence and multi-agent systems.
Uncertainty theories: an introduction (presentation)
The past decades have viewed a growing interest for models of uncertainty that go beyond the classical models that are set/intervals and probabilities. In this tutorial, I will introduce some basic ideas about uncertainty, browsing different general situations and problems where uncertainty can exist. I will then provide an overview of the different theories aiming at modelling such uncertainty and at reconciling the two well-known models that are intervals and sets on one side, probabilities on the other side. Finally, I will briefly address the problems of inference and decision through some basic examples. Although I will occasionally touch upon some fundamental questions, the view adopted in the tutorial will be a pragmatic one.
We originally planned for a workshop in addition to the tutorial program; in the end we have decided against this. There will therefore be no workshop.