TBA
On monday, at 2pm - UPEC CMC - Room P2-131
December 18, 2023
TBA
TBA
December 11, 2023
TBA
TBA
December 4, 2023
Timed Automata Verification and Synthesis via Finite Automata Learning
We present algorithms for model checking and controller synthesis of timed automata, seeing a timed automaton model as a parallel composition of a large finite-state machine and a relatively smaller timed automaton, and using compositional reasoning on this composition. We use automata learning algorithms to learn finite automata approximations of the timed automaton component, in order to reduce the problem at hand to finite-state model checking or to finite-state controller synthesis. We present an experimental evaluation of our approach.
November 27, 2023
Learning the Structure of Bayesian Networks using Constraint Programming
Bayesian networks are probabilistic graphical models with a wide range of application areas including gene regulatory network inference, risk analysis and image processing. Learning the structure of a Bayesian network (BNSL) from discrete data is known to be an NP-hard task with a superexponential search space of directed acyclic graphs.
In this talk, I will present recent work on solving the BNSL using constraint programming. Building on previous work on solving BNSL with CP and with ILP, we propose new algorithms for handling the difficult acyclicity constraint and the so-called cluster cuts that can be generated from it. We give a new polynomial time algorithm for discovering a subset of all possible cluster cuts, a greedy algorithm for approximately solving the resulting linear program, and a generalised arc consistency algorithm for the acyclicity constraint. These improve performance by orders of magnitude compared to previous CP-based approaches, but scalability is still limited by the fact that the basic decision variables used in this model have domain size Omega(n^log n). We propose a novel representation of domains using decision trees and show that relatively simple operations are sufficient to implement all propagation and bounding algorithms. The combination of the algorithmic techniques and the new representation result in a solver which compares favourably with all competing state-of-the-art approaches.
November 20, 2023
Stochastic Modeling and Optimization For Power and Performance Control In DVFS Systems
This work addresses the problem of performance-energy trade-off in DVFS (Dynamic Voltage Frequency Scaling) systems. We propose a stochastic hybrid model between hysteresis models and server block models. We provide a closed form for the steady-state distribution probability and we establish a “st” type order to compare the performance measures.
The fast computation of power and performance measures leads to a multi-objective optimization analysis in two forms: a scalarization method and a Pareto based method. For the two approaches, we propose fast and efficient approximate algorithms that construct progressively an optimal solution. To discuss results, the model is used to simulate a physical server hosting several VMs (Virtual Machines) where we investigate optimal thresholds for the performance-energy trade-off.
November 13, 2023
Automates communicants et communications quasi-synchrones
Les systèmes distribués sont le plus souvent basés sur l’échange asynchrone de messages entre des agents. Les automates communicants sont un formalisme permettant de modéliser les communications de ces systèmes, afin de détecter automatiquement des erreurs comme des pertes de messages ou des inter-blocages. Ces problèmes sont indécidables en général pour des systèmes à partir de deux machines, et plusieurs hypothèses restrictives ont été étudiées pour les rendre décidables. Dans cette présentation, nous étudierons une de ces approches, basée sur l’étude des systèmes dont les exécutions sont réalisables avec des communications synchrones (Realisable with Synchronous Communication, ou RSC). Les comportements de ces systèmes approximent des comportements synchrones, où les messages sont envoyés et reçus simultanément.
October 16, 2023
Inside and Beyond Decisiveness for Infinite Probabilistic Models
Decisiveness of infinite Markov chains is a key property that allows to compute reachability probabilities up to an arbitrary precision. However such a generic method suffers from two limitations: (1) decisiveness is somewhat related to recurrence of Markov chains and so does not cover transient models and (2) most of the applicable cases of decisiveness require that the transition weights are independent of the current state which forbids a relevant fraction of probabilistic modelings requiring dynamic weights (i.e., also depending on the current state). In this work, we extend this approach in two directions. First we introduce a new property called divergence that somehow complements decisiveness and also leads to a similar computation. Then we study the decidability of these two properties for dynamic probabilistic versions of counter machines, pushdown automata, and Petri nets. Finally, we exhibit some divergent or decisive subclasses of channel systems and Petri nets.
This presentation is based on two papers accepted by CONCUR’23 and by RP’23, and are joint work with Alain Finkel (LMF) and Serge Haddad (LMF).
October 9, 2023
Queens of the Hill
Inspired by the programming game Core Wars, we propose in this work a framework and the organisation of king of the hill-style tournaments between P systems. We call these tournaments Queens of the Hill and the individual contestants valkyries. The goal of each valkyrie is to dissolve as many membranes of as many other valkyries as possible, while at the same time resisting the attacks. Valkyries are transition P systems with cooperative rules, target indication, and rudimentary matter–anti-matter annihilation rules. These ingredients are sufficient for computational completeness, but the context of Queens of the Hill reduces the relevance of this statement. We give some tentative examples of strategies and discuss their advantages and drawbacks. Finally, we describe how Queens of the Hill can be used as a teaching exercise, and also a tool to federate the students’ creativity to push the frontiers of membrane computing.
October 2, 2023
Génération automatique de code parallèle isochrone
Depuis la stagnation de la fréquence d’horloge des processeurs, l’accroissement de la puissance de calcul a dépendu entièrement de l’accroissement du nombre d’unités de calcul.
Plus que la difficulté algorithmique impliquée par l’écriture de tout programme séquentiel, la programmation parallèle demande au programmeur de gérer de nombreuses unités de calcul, incluant leurs tâches et leurs interactions.
Pour alléger le fardeau du programmeur, cette présentation propose deux approches différentes de génération automatique de code parallèle.
Le modèle parallèle isochrone BSP possède des propriétés intéressantes telles que sa simplicité et son modèle de coût qui en font la cible de notre génération de code parallèle.
Les automates et expressions régulières sont souvent choisis pour modéliser les calculs séquentiels et leurs parallélisation devrait, à long terme, aboutir à de solide fondations pour la génération de code parallèle.
Pour notre approche principale, nous développons la théorie des automates BSP avec leur génération et déterminisation.
Cette théorie est utilisée dans une nouvelle méthode pour la recherche parallèle de motif à l’aide d’expressions régulières.
Notre autre approche propose un langage spécifique au domaine des réseaux de neurones où la composition fonctionnelle d’un petit nombre de primitives facilite le développement, la maintenance et la définition formelle du langage par rapport aux approches existantes.