JELIA 2010 will feature invited talks by the following world-leading researchers:
- Gerhard Brewka (University of Leipzig, Germany)
- Adnan Darwiche (UCLA, USA)
- Stephane Demri (Centre National de la Recherche Scientifique CNRS, France)
The abstracts and schedule of the invited talks are provided below.
Relax, Compensate and then Recover:
A Theory of Anytime, Approximate
A Theory of Anytime, Approximate Inference
I will present in this talk a theory of anytime, approximate inference, which explains some of the most influential algorithms in probabilistic reasoning, yet is based on one of the older ideas in symbolic reasoning: Relaxations. According to this theory, the fundamental notion of approximation is that of "relaxing" logical constraints (equalities in particular) for the purpose of decomposing a problem into smaller pieces that can be solved independently. A second fundamental notion is that of "compensation", which calls for imposing weaker and, typically, probabilistic notions of equality to compensate for the relaxed equalities. The third fundamental notion of the presented theory is that of "recovery," where some of the relaxed equalities are incrementally recovered, based on an assessment of their impact on improving the quality of approximations. I will discuss how this theory subsumes one of the most influential algorithms in probabilistic inference: loopy belief propagation and some of its generalizations. I will also discuss the relationship between this theory and current developments in symbolic reasoning -- in particular, how advances in knowledge compilation can impact the state of the art in approximate (and exact) probabilistic inference.
Counter Systems for Data Logics
Data logics are logical formalisms that are used to specify properties on structures equipped with data (data words, data trees, runs from counter systems, timed words, etc.). In this survey talk, we shall see how satisfiability problems for such data logics are related to reachability problems for counter systems (including counter automata with errors, vector addition systems with states, etc.). This is the opportunity to provide an overview about the relationships between data logics and verification problems for counter systems.
Nonmonotonic Tools for Argumentation
In argumentation, the question whether a (pro)position p is accepted or not is decided by constructing and possibly weighing arguments pro and con p. The arguments are generated from a - possibly inconsistent and/or defeasible - knowledge base. Several nonmonotonic tools have been developed which abstract away from the specific content of the arguments and focus on particular services, for instance for conflict solving among arguments.
The most popular example of such a tool are Dung's argumentation frameworks (AFs) which define a "calculus of opposition" and are probably the simplest nonmonotonic systems available. Given a set of arguments with an attack relation among them, AFs come with different semantics, where a semantics specifies which subsets of the arguments are acceptable. The different semantics represent different intuitions how to select arguments, as well as different degrees of skepticism.
In the talk I will present abstract dialectical frameworks (ADFs), a powerful generalization of AFs where nodes in the argument graph, rather than having an implicit acceptance condition, come with an explicit boolean function specifying when the node is to be accepted based on the status of its parents. This allows us to represent support in addition to attack, and to express flexible ways of taking pro and con arguments into account.
We illustrate the usefulness of ADFs by reconstructing Carneades argument evaluation structures (CAES). Carneades, developed by Gordon, Prakken and Walton, is an influential and widely cited argumentation system, motivated by the needs of legal argumentation. It covers relevant aspects such as burden of proof, proof standards and the like. We show how CAES can be reconstructed as ADFs. This not only demonstrates the generality of ADFs. It also allows us to lift a restriction of CAES to acyclic sets of arguments and provides the generalized systems with the standard semantics developed