Conference Tutorials
|
Laszlo Gulyas, PhD dir. of research AITIA International Inc. Czetz János u. 48-50., Budapest 1039, Hungary Assistant professor Department of History and Philosophy of Science, Lorand Eotvos University, Budapest.  Tel: +36 1 453 8080 Fax: +36 1 453 8081 Email: lgulyas@aitia.ai
|
Agent-Based Modeling and Simulation with the Multi-Agent Simulation Suite
Agent-based modeling and simulation is a novel computational approach to model complex social systems. Its main tenet is to model the individual, together with its imperfections (e.g., limited cognitive and computational abilities), its idiosyncrasies and unique interactions. Thus, the approach builds the model from the bottom-up, focusing mostly on micro rules and seeking the understanding of the emergence of macro behavior.
The tutorial will introduce the participants to this novel method, via worked examples using the Multi-Agent Simulation Suite (MASS). The discussed topics will range from model creation and definition (using the FABLES language), via model exploration and experimental design (using MEME, the Model Exploration Module and clusters of computers or grids), to results analysis (using MEME and other tools). The models will be web-enabled and their participatory version, mixing human controlled agents into the system, will also be created (using the Participatory Extension, PET). Discussions about the potential uses and benefits of experimenting with such hybrid systems will close the day.
Biography
László Gulyás (Ph.D. in Computer Science) is assistant professor at the Department of History and Philosophy of Science, Lorand Eotvos University, Budapest. He is also a research partner at AITIA International Inc and a fellow at Collegium Budapest (Institute for Advanced Study). He is a member of the Scientific Advisory Board of the Simulation Center of the Informatics Cooperative Research and Education Center of the Eötvös Loránd University. He spent three semesters at Harvard University's Government Department and at the Center for Basic Research in the Social Sciences (CBRSS) as a research associate.
He has been doing research on agent-based modeling and multi-agent systems since 1996. He leads the development of the Multi-Agent Simulation Suite (MASS) and the Functional Agent-Based Language for Simulations (FABLES) In the past, he has led the development of the Multi-Agent Modeling Language (MAML), the first special purpose programming language for agent-based simulation. He also contributed to the design and development of RePast, one of the leading second generation agent-based simulation environments. He's been involved in teaching both graduate and undergraduate level courses in agent-based modeling and simulation at Harvard University, at the Central-European University and at the Eötvös Loránd University, Hungary. He co-directed the Complex Systems and Social Simulation summer school at the Central European University's Summer University, Budapest, 2008, and was also a faculty member at the 2002 Budapest Complex Systems Summer School organized by the Santa Fe Institute. Dr. Gulyás has authored several book chapters (5+) and journal articles (5+), and published many conference papers (40+). He participated in two international research consortia under the European Commission's 6th Framework Programme and was project leader or participant in 4 R&D projects funded by the Hungarian Government.
Dr. Gulyás is a graduate of the Eötvös Loránd University, Hungary, from where he received his PhD, MSc and BSc degrees, all in Computer Science.
His main research interests are computational multi-agent systems where he has worked on 'engineering' desired emergent phenomena. He is currently working on agent-based models of social systems.
|
Michel Cotsaftis |
What makes a system complex? An approach to self-organization and emergence
The fast changing reality in technical and natural domains perceived by always more accurate observations has drawn the attention on a new and very broad class of systems mainly characterized by specific behavior which has been entered under the common wording "complexity". Based on elementary system graph representation with components as nodes and interactions as vertices, it is shown that systems belong to only three states: simple, complicated, and complex, the main properties of which are discussed. The first two states have been studied at length over past centuries, and the last one finds its origin in the elementary fact that when system performance is pushed up, there exists a threshold above which interaction between components overtake outside interaction. At the same time, system self-organizes and filters corresponding outer action, making it more robust to outer effect, with emergence of a new behavior which was not predictable from only components study. Examples in Physics and Biology are given, and three main classes of "complexity" behavior are distinguished corresponding to different levels of difficulty to handle the problem of their dynamics. The great interest of using complex state properties in man-made systems is stressed and important issues are discussed. They mainly concentrate on the difficult balance to be established between the relative system isolation when becoming complex and the delegation of corresponding new capability from (outside)operator. This implies giving the system some "intelligence" in an adequate frame between the new augmented system state and supervising operator, with consequences on the canonical system triplet {effector-sensor-controller} which has to be reorganized in this new setting. Moreover, it is observed that entering complexity state opens the possibility for the function to feedback onto the structure, i.e. to mimic at technical level the invention of Nature over Her very long history.
|
Prof. Jean-Pierre Müller |
Agent-based modelling with ontologies and dynamical behavioral systems
A model is a mean to answer questions about a given system. Therefore, it depends both on the question and on the conceptual background of the scientist. The conceptual background entirely defines how the scientist is looking at the system. Most modeling platforms are only dedicated to defining the model at a more or less high level of abstraction and running it . If one want to support the whole modeling process, the formulation of the conceptual background using ontologies becomes necessary, especially when more than one conceptual backgrounds from a variety of scientists have to be articulated. The aim of this tutorial is to introduce Mimosa, a modeling and simulation platform. The modeling process in Mimosa is organized in two stages presented in two sessions. The first stage consists in defining the conceptual and concrete model using ontology(ies). The notion of ontology and how it is implemented shall be described and experimented on use cases. The second stage consists in defining the mapping of the ontology(ies) within a model structure and its dynamics. This definition is made through ontology annotation which shall be introduced and illustrated on examples.
Biography
Jean-Pierre Müller's research topic is the modeling and simulation of complex systems with a focus towards the emergence of individual and collective behaviour. The long term goal is to elaborate a theoretical foundation to or at least an understanding of human knowledge and behavior. Jean-Pierre Müller has been professor at the University of Neuchâtel where it managed the CASCAD Team. He has been animator of the COLLINE (COLLectif-INteraction-Emergence) working group of AFIA and GDR-I3 and responsible for the Working group SMA (Multi-Agent Systems) of GDR-I3. He is currently managing the team GREEN (Common Pool Ressources Management) whose aim is to support collective decision making by modelling and simulation using, among others, multi-agent systems. Jean-Pierre Müller is co-editor of 5 books and author and co-author of more than 50 publications. It is also associate researcher to LIRMM.
Design and Analysis of Computer Experiments with SimExplorer
Thierry Faure, Nicolas Dumoulin, Florent Chu art, Guillaume De uant, Cemagref, Aubiere, France
|
The attendees will acquire the concepts and some practice of SimExplorer, allowing them to use it autonomously for their own needs. SimExplorer is a frame-work that includes an API1 and an IDE2 dedicated to design and analysis of computer experiments. SimExplorer combines blind computing on demand service access, traceability and reporting features and sensibility analysis tools. This tutorial illustrates a typical use of SimExplorer, performing a global sensitivity analysis of a given individual-based predator prey model. |
You can download the extend tutorial description here. 
Multi-Agent Systems Implementations – Practical Sessions
|
|
Basic learning of RePast and presentation of RePastJ, RePastPy and RePast Symphony. We will focus on two specific aspects of Repast: the mixing MAS-GIS, using RePastJ and the extensions proposed by AITIA International Inc : FABLES, MEME and other MASS (Multi-Agent Simulation Suite) tools
Instructors and organizers
Laszlo Gulyas, Cyrille Bertelle and Rawan Ghnemat
1 full day
Tutorial on Mimosa. A first session will be devoted to modelling using ontologies with Mimosa. A second session will be devoted to dynamical aspects with Mimosa
Instructor
Jean-Pierre Müller
1 full day
Graphstream
Basic tutorial of GraphStream which is a java library that manages dynamic graphs. It is composed of an object oriented API that gives an easy and quick way to construct and evolve edges and nodes in a graph
Instructors
Antoine Dutot and Yoann Pigné
Half a day
