AI*IA 2003 - Ottavo Congresso Nazionale
dell'Associazione Italiana per l'Intelligenza Artificiale
23-26 Settembre 2003, Polo didattico "L. Fibonacci", Universitą di Pisa
TUTORIAL T1 (23 Settembre, 9:30 - 13:30)
TITLE: An introduction to logic-based multi-agent systems.
ABSTRACT: Multi-agent systems study and development is a lively research area which integrates aspects of Artificial Intelligence with some others of Distributed Systems. The idea is to specify, describe, analyse, verify and implement systems based on multiple intelligent entities called agents.
While consolidated artifical intelligence techniques such as logic based deductive reasoning, hypothetical reasoning, planning and learning lend themselves well to the development of intelligent agents, on the other hand, a formal ground in the specification of agent systems is what is needed when the purpose is to guarantee and verify properties.
The aim of this tutorial is to give a general overview on the state of the art of logic-based multi-agent systems research, with particular respect to computational logic. We will present some reference architectures and languages where logic plays an important role, and we will discuss advantages and limitations of a logic based approach to agent system design and implementation.
Being this an introductory tutorial, no background knowledge of any kind is presumed.
TUTORIAL T2 (23 Settembre, 9:30 - 13:30)
TITLE: Machine Learning for Automated Text Classification
ABSTRACT: The automated categorization (or classification) of texts into prespecified categories, although dating back to the early '60s, has witnessed a booming interest in the last ten years, due to the increased availability of documents in digital form and the ensuing need to organize them. In the research community the dominant approach to this problem is based on the application of supervised machine learning techniques: a general inductive process automatically builds a classifier by learning, from a set of previously classified documents, the characteristics of one or more categories. The advantages of this approach over the knowledge engineering approach (consisting in the manual definition of a classifier by domain experts) are a very good effectiveness, considerable savings in terms of expert manpower, and straightforward portability to different domains. In this survey we look at the main approaches that have been taken towards automatic text categorization within the machine learning paradigm. We will discuss in detail issues pertaining to three different problems, namely document representation, classifier learning, and classifier evaluation.
The objective of this tutorial is to make the attendees aware of the concepts and techniques for automatically or semiautomatically classifying documents into a set of topical categories, possibly organized into a hierar chical structure. The course is addressed to students and researchers active in natural language processing or in machine learning and interested in the application of qualitative and (mostly) quantitative techniques for automatically dealing with large corpora of texts.
The level is introductory. No prerequisites are needed, apart from a generic knowledge of maths and computer science fundamentals; basic concepts of information retrieval and machine learning will be explicitly introduced as they are needed.
TUTORIAL T3 (23 Settembre, 14:30 - 18:30)
TITLE: Norms and Institutions
ABSTRACT: This will be an overview of developments in formal-logical methods for modelling patterns of agent interaction in organisations and societies, and of the rules and policies governing such interactions. The focus of attention is on agents' obligations, permissions, and rights, their responsibilities in regard to the achievement of specific goals, and their powers to initiate changes, to authorise, and to delegate. The term "agents" is intended to cover both human and artificial (computer) agents. The presentation will be based on illustrative examples, in the formal representation of duties and rights, in the specification of computer systems, and in the modelling of organisations and computational societies.
Technical developments will be outlined, but the presentation will not presuppose familiarity with more than elementary propositional logic.
TUTORIAL T4 (23 Settembre, 14:30 - 18:30)
TITLE: Fusion of Multiple Patterns Classifiers
ABSTRACT: In the field of pattern recognition, fusion of multiople classifiers is currently used for solving difficult recognition tasks and designing high performance systems. From a theoretical viewpoint, fusion of multiple classifiers allows overcoming the well known limitations of classical approaches to design a pattern recognition system that focises on the search of the best individual classifier. From a practical viewpoint, the concept of multiple classifiers derives naturally from the context and requirements of many applications. As an example, in applications dealing with multiple sensor systems, theory of multiple classifiers fits well with the need of decigning decision fusion modules based on a variety of sensor types. This tutorial will illustrate the theoretical foundations of fusion of multiple classifiers and present the main methods and algorithms. We will show how fusion of multiple classifiers can be used for designing high performance pattern recognition and decision fusion systems. Applications of multiple classifiers dealing with security systems (e.g., biometric, video-surveillance, intrusion detection in computer networks) will be presented.
No previous knowledge of the tutorial topics will be assumed. However, some basic familiarity with pattern recognition theory is helpful.