ENS Rennes


ENS Rennes

Time filter
Source Type

Berthon R.,ENS Rennes | Randour M.,Free University of Colombia | Raskin J.-F.,Free University of Colombia
Leibniz International Proceedings in Informatics, LIPIcs | Year: 2017

The beyond worst-case synthesis problem was introduced recently by Bruyère et al. [10]: it aims at building system controllers that provide strict worst-case performance guarantees against an antagonistic environment while ensuring higher expected performance against a stochastic model of the environment. Our work extends the framework of [10] and follow-up papers, which focused on quantitative objectives, by addressing the case of ω-regular conditions encoded as parity objectives, a natural way to represent functional requirements of systems. We build strategies that satisfy a main parity objective on all plays, while ensuring a secondary one with sufficient probability. This setting raises new challenges in comparison to quantitative objectives, as one cannot easily mix different strategies without endangering the functional properties of the system. We establish that, for all variants of this problem, deciding the existence of a strategy lies in NP \ coNP, the same complexity class as classical parity games. Hence, our framework provides additional modeling power while staying in the same complexity class. © Raphaël Berthon, Mickael Randour, and Jean-François Raskin;.

Hunter A.,British Columbia Institute of Technology | Schwarzentruber F.,ENS Rennes
CEUR Workshop Proceedings | Year: 2015

Public announcements cause each agent in a group to modify their beliefs to incorporate some new piece of information, while simultaneously being aware that all other agents are doing the same. Given some fixed goal formula, it is natural to ask if there exists an announcement that will make the formula true in a multi-agent context. This problem is known to be undecidable in a general modal setting, where the presence of nested beliefs can lead to complex dynamics. In this paper, we consider not necessarily truthful public announcements in the setting of propositional belief revision. We are given a goal formula for each agent, and we are interested in finding a single announcement that will make each agent believe the corresponding goal following AGM-style belief revision. If the goals are inconsistent, then this can be seen as a form of ampliative reasoning. We prove that determining if there is an arbitrary public announcement in this setting is not only decidable, but that it is simpler than the corresponding problem in the most simplified modal logics. Moreover, we argue that propositional announcements and beliefs are sufficient for modelling many practical problems, including simple robot controllers.

Schwarzentruber F.,ENS Rennes | Semmling C.,Ruhr University Bochum
Frontiers in Artificial Intelligence and Applications | Year: 2014

STIT is a potential logical framework to capture responsibility, counterfactual emotions and norms, which are main ingredients for specifying behaviors of virtual agents. We identify here a new fragment and its satisfiability problem is NP-complete and in Σ3 when the number of agents is unbounded. We also identify a slightly more expressive fragment which is undecidable. © 2014 The Authors and IOS Press.

Van Eijck J.,CWI and ILLC | Schwarzentruber F.,ENS Rennes
Advances in Modal Logic | Year: 2014

We propose a simplified logic for reasoning about (multi-agent) epistemic probability models, and for epistemic probabilistic model checking. Epistemic probability models are multi-agent Kripke models that assign to each agent an equivalence relation on worlds, together with a function from worlds to positive rationals (a lottery). The difference with the usual approach is that probability is linked to knowledge rather than belief, and that knowledge is equated with certainty. A first contribution of the paper is a comparison of a semantics for epistemic probability in terms of models with multiple lotteries and models with a single lottery. We give a proof that multiple lottery models can always be replaced by single lottery models. As multiple lotteries represent multiple subjective probabilities, our result connects subjective and intersubjective probability. Next, we define an appropriate notion of bisimulation, and use it to prove an adaptation of the Hennessy-Milner Theorem and to prove that some finite multiple lottery models only have infinite single lottery counterparts. We then prove completeness, and state results about model checking complexity. In particular, we show the PSPACE-completeness of the model checking in the dynamic version with action models. The logic is designed with model checking for epistemic probability logic in mind; a prototype model checker for it exists. This program can be used to keep track of information flow about aleatory acts among multiple agents.

Royer A.,ENS Rennes | Lampert C.H.,IST Austria
Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition | Year: 2015

Classifiers for object categorization are usually evaluated by their accuracy on a set of i.i.d. test examples. This provides us with an estimate of the expected error when applying the classifiers to a single new image. In real application, however, classifiers are rarely only used for a single image and then discarded. Instead, they are applied sequentially to many images, and these are typically not i.i.d. samples from a fixed data distribution, but they carry dependencies and their class distribution varies over time. In this work, we argue that the phenomenon of correlated data at prediction time is not a nuisance, but a blessing in disguise. We describe a probabilistic method for adapting classifiers at prediction time without having to retrain them. We also introduce a framework for creating realistically distributed image sequences, which offers a way to benchmark classifier adaptation methods, such as the one we propose. Experiments on the ILSVRC2010 and ILSVRC2012 datasets show that adapting object classification systems at prediction time can significantly reduce their error rate, even with no additional human feedback. © 2015 IEEE.

Grossi D.,University of Liverpool | Lorini E.,University Paul Sabatier | Schwarzentruber F.,ENS Rennes
Journal of Artificial Intelligence Research | Year: 2015

The article introduces a ceteris paribus modal logic, called CP, interpreted on the equivalence classes induced by finite sets of propositional atoms. This logic is studied and then used to embed three logics of strategic interaction, namely atemporal STIT, the coalition logic of propositional control (CL-PC) and the starless fragment of the dynamic logic of propositional assignments (DL-PA). The embeddings highlight a common ceteris paribus structure underpinning the key operators of all these apparently very different logics and show, we argue, remarkable similarities behind some of the most influential formalisms for reasoning about strategic interaction. © 2015 AI Access Foundation. All rights reserved.

Cheriere N.,ENS Rennes | Saule E.,University of North Carolina at Charlotte
Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2015 | Year: 2015

When the size of parallel systems increases, centralized algorithms to schedule tasks on the system can induce a significant overhead. This is why decentralized scheduling algorithms have been developed. The most popular one certainly is work-stealing because of its interesting theoretical guarantees. Parallel systems have evolved from homogeneous clusters to fully heterogeneous ones such as GPU-accelerated clusters. We investigate in this paper decentralized scheduling algorithms for heterogeneous systems. The guarantees of work-stealing algorithms no longer hold on such systems because it is an a posteriori algorithm which highly depends on the initial distribution of work. We focus on a priori decentralized scheduling algorithms for heterogeneous systems and we propose two distributed algorithms to balance the load on unrelated machines for two particular cases. The first one exploits a low heterogeneity in the task set and reaches an approximation ratio linear in the number of types of tasks. The second one focuses on the case where the system only uses two different types of machines and we show it is a 2-approximation if the system converges. In the case it does not converge, we study the dynamic equilibrium of the system. In the homogeneous case, we numerically compute the probability density function of the load imbalance and show that the imbalance is low on average. And we show using simulation that the heterogeneous case is similar to the homogeneous case and that the imbalance is low in both cases. © 2015 IEEE.

Berthon R.,ENS Rennes | Ringeissen C.,French Institute for Research in Computer Science and Automation
CEUR Workshop Proceedings | Year: 2016

Free Data Structures are finite semantic trees modulo equational axioms that are useful to represent classical data structures such as lists, multisets and sets. We study the satisfiability problem when free data structures are combined with bridging functions. We discuss the possibility to get a combination method à la Nelson-Oppen for these particular non-disjoint unions of theories. In order to handle satisfiability problems with disequalities, we investigate a form of sufficient surjectivity for the bridging functions.

Charrier T.,ENS Rennes | Schwarzentruber F.,ENS Rennes
Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS | Year: 2015

We propose a variant of arbitrary public announcement logic which is decidable. In this variant, knowledge accessibility relations are defined by programs. Technically, programs are written in dynamic logic with propositional assignments. We prove that both the model checking problem and the satisfiability problem are decidable and AEXpol-complete where AEXpol is the class of decision problems decided by alternating Turing machines running in exponential time where the number of alternations is polynomial. Whereas arbitrary public announcement logic is undecidable, our framework is decidable and we provide a proof-of-concept to show its expressiveness: we use our framework to reason about epistemic properties and arbitrary announcements when agents are cameras located in the plane. Copyright © 2015, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.

Schwarzentruber F.,ENS Rennes
Journal of Logic, Language and Information | Year: 2015

This paper describes methods for generating interactive Euler diagrams. User interaction is needed to improve the aesthetic quality of the drawing without writing tedious formal specifications. More precisely, the user can modify the diagram’s layout on the fly by mouse control. We prove that the satisfiability problem is in PSPACE and we provide two syntactic fragments such that the corresponding restricted satisfiability problem is already NP-hard. We describe (1) an improved local search based approach, (2) a method inspired from the gradient method and a hybrid method mixing both (1) and (2). A software tool was implemented and its implementation is described. We also experimentally compare the different methods. We first see that the improved local search and the hybrid method outperforms the local search from the literature and the gradient method for generating a diagram. Concerning interaction, the local search approach is not suitable but hybrid method and gradient method give both good results in terms of quality of drawings and stability. Specifications are written using region connection calculus (RCC-8), radius constraints and disjunctions. Euler diagrams are described as set of circles. © 2015, Springer Science+Business Media Dordrecht.

Loading ENS Rennes collaborators
Loading ENS Rennes collaborators