Search

Item does not contain fulltext

Item does not contain fulltext

Logic-based preference representation languages are promising for expressing preferences over combinatorial domains. Sets of weighted formulas, called goalbases, can be used to define several such languages. How goalbases are translated into utility functions---that is, by what aggregation function this is done---is a crucial component of this type of language. In this paper, we consider the properties of several goalbase languages which use max as their aggregation function. In particular, we examine the expressivity, succinctness and complexity of such languages.

We describe a generalized Q-learning type algorithm for reinforcement learning in competitive multi-agent games. We make the observation that in a competitive setting with adaptive agents an agent's actions will (likely) result in changes in the opponents policies. In addition to accounting for the estimated policies of the opponents, our algorithm also adjusts these future opponent policies by incorporating estimates of how the opponents change their policy as a reaction to ones own actions. We present results showing that agents that learn with this algorithm can successfully achieve high reward in competitive multi-agent games where myopic self-interested behavior conflicts with the long term individual interests of the players.We show that this approach successfully scales for multi-agent games of various sizes, in particular to the social dilemma type problems: from the small iterated Prisoner's Dilemma, to larger settings akin to Harding's Tragedy of the Commons. Thus, our multi-agent reinforcement algorithm is foresighted enough to correctly anticipate future rewards in the important problem class of social dilemmas, without having to resort to negotiation-like protocols or precoded strategies.

International audience

The aim of this article is to provide a logic which supports reasoning about institutional dynamics. In our approach an institution is grounded on the acceptances of its members, and its dynamics depend on the dynamics of these acceptances. We describe the power of the role legislator to create and modify the legal level of an institution. We will also provide an analysis of two important concepts in social theory: constitutive rule and norm of competence.

The logic FO(ID) uses ideas from the field of logic programming to extend first order logic with non-monotone inductive definitions. This paper studies a deductive inference method for PC(ID), its propositional fragment. We introduce a formal proof system based on the sequent calculus (Gentzen-style deductive system) for this logic. As PC(ID) is an integration of classical prepositional logic and propositional inductive definitions, our deductive system integrates inference rules for propositional calculus and definitions. We prove the soundness and completeness of this deductive system for a slightly restricted fragment of PC(ID). We also give a counter-example to show that cut-elimination does not hold in this proof system. © Springer-Verlag Berlin Heidelberg 2007. ispartof: pages:162-174 ispartof: Lecture Notes in Computer Science vol:4483 pages:162-174 ispartof: Logic Programming and Nonmonotonic Reasoning location:Tempe, AZ, USA date:15 May - 17 May 2007 status: published

International audience; This work proposes a formal modelling of categorisation processes attempting at simulating the way information is categorised by neural populations in the human brain. The formalism mainly relies on a similarity-based approach to categorisation. It involves weighted rules that use inference and fusion techniques borrowed from fuzzy logic. The approach is illustrated by a simulation of the McGurck effect where the combination of contradictory auditory and visual stimuli creates an auditory perceptive illusion.

The fundamental operation of dominance testing, i.e., determining if one alternative is preferred to another, is in general very hard for methods of reasoning with qualitative conditional preferences such as CP-nets and conditional preference theories (CP-theories). It is therefore natural to consider approximations of preference, and upper approximations are of particular interest, since they can be used within a constraint optimisation algorithm to find some of the optimal solutions. Upper approximations for preference in CP-theories have previously been suggested, but they require consistency, as well as strong acyclicity conditions on the variables. We define an upper approximation of conditional preference for which dominance checking is efficient, and which can be applied very generally for CP-theories.

This paper presents a new Abductive Logic Programming (ALP) approach for assisting clinicians in the selection of antiretroviral drugs for patients infected with Human Immunodeficiency Virus (HIV). The approach is comparable to laboratory genotypic resistance testing in that it aims to determine which viral mutations a patient is carrying and predict which drugs they are most likely resistant to. But, instead of genetically analysing samples of the virus taken from patients-which is not always practicable-our approach infers likely mutations using the patient's full clinical history and a model of drug resistance maintained by a leading HIV research agency. Unlike previous applications of abduction, our approach does not attempt to find the "best" explanations, as we can never be absolutely sure which mutations a patient is carrying. Rather, the intrinsic uncertainty of this domain means that multiple alternative explanations are inevitable and we must seek ways to extract useful information from them. The computational and pragmatic issues raised by this approach have led us to develop a new ALP methodology for handling numerous explanations and for drawing predictions with associated levels of confidence. We present our in-Silico Sequencing System (iS3) for reasoning about HIV drug resistance as a concrete example of this approach. © 2006 The authors. 141 437 441

In this paper we show how to extend clausal temporal resolution to the ground eventuality fragment of monodic first-order temporal logic, which has recently been introduced by Hodkinson, Wolter and Zakharyaschev. While a finite Hilbert-like axiomatization of complete monodic first order temporal logic was developed by Wolter and Zakharyaschev, we propose a temporal resolutionbased proof system which reduces the satisfiability problem for ground eventuality monodic first-order temporal formulae to the satisfiability problem for formulae of classical first-order logic.

In this paper, we study how several patterns of deductive generalization from positive and negative examples can be relaxed to handle forms of defeasible reasoning, using default logic as a case study. We compare the resulting paradigms and establish the logical conditions under which they can take place.