Papers by Michel Liquiere
HAL (Le Centre pour la Communication Scientifique Directe), 2008
Ce travail porte sur la mise en evidence par apprentissage de regularites dans la description d&#... more Ce travail porte sur la mise en evidence par apprentissage de regularites dans la description d'objets complexes. Pour la description des exemples, nous utilisons un formalisme objet base sur le modele des graphes conceptuels de sowa. La recherche de regularites valides est ramenee a la recherche de sous-graphes conceptuels, donc connexes, apparaissant souvent dans les graphes decrivant les exemples et rarement dans ceux decrivant les contre-exemples. Nous proposons un algorithme qui procede en deux etapes: 1) recherche de chemins conceptuels; 2) assemblage de ces chemins pour former des arbres conceptuels. Les regularites extraites par cette methode sont ensuite structurees par l'intermediaire d'un treillis de galois, ce qui permet de batir des methodes tres interessantes de decision et d'explication des resultats. Ces algorithmes ont ete exploites dans le domaine de la biologie sur des problemes de grandes tailles
Intellectica. Revue de l'Association pour la Recherche Cognitive, 1987
Etude et réalisation de mécanismes d'argumentations de décisions . Les connaissances étant ex... more Etude et réalisation de mécanismes d'argumentations de décisions . Les connaissances étant extraites è partir d'exemples. Application è un problème d'éthologie.
Résumé : Nous discutons dans cet article les bases d'un projet scientifique commun aux in-fo... more Résumé : Nous discutons dans cet article les bases d'un projet scientifique commun aux in-formaticiens, cogniticiens et économistes, dont le but est de modéliser et d'expérimenter des situations de recherche collective en science. Nous avons développé une plateforme ...
SIGLEINIST T 75865 / INIST-CNRS - Institut de l'Information Scientifique et TechniqueFRFranc
New Approaches in Classification and Data Analysis, 1994
This work describes a new polynomial algorithm for structural learning. This algorithm is special... more This work describes a new polynomial algorithm for structural learning. This algorithm is specially design to deal with: large amounts of objects, object linked to many other objects, noisy data.
Applied Artificial Intelligence, 2003
... For their contribution to the reviewing process of both special issues on CLKDD, we would lik... more ... For their contribution to the reviewing process of both special issues on CLKDD, we would like to thank Artur Bykowski, Claudio Carpineto, Peter Eklund, Edwin Diday, Bernhard Ganter, Robert Godin, Derrick DG Kourie, Sergei O. Kuznetsov, Lotfi Lakhal, Rokia Missaoui ...
This work is subject to copyright. All rights reserved. Reproduction or publication of this mater... more This work is subject to copyright. All rights reserved. Reproduction or publication of this material, even partial, is allowed only with the editors ’ permission. Technical Editor:
International Journal of General Systems
Communications in Computer and Information Science, 2011
In this paper, we tackled the problem of generation of rare classification rules. Our work is mot... more In this paper, we tackled the problem of generation of rare classification rules. Our work is motivated by the search of an effective algorithm allowing the extraction of rare classification rules by avoiding the generation of a large number of patterns at reduced time. Within this framework we are interested in rules of the form a 1 ∧ a 2… ∧ a n ⇒b
Abstract. In this article, we outline some issues in agent interaction on the Web, which is the c... more Abstract. In this article, we outline some issues in agent interaction on the Web, which is the center point of supporting the needs of fully-realized learning GRID in the future. Of particular importance is conversation support, with its core element, communication protocols. We propose to construct communication protocols through concept learning of services generated on the Web. The proposed approach incorporates a machine learning model into a conversation environment for the induction of communication protocols.
International Journal on Artificial Intelligence Tools, 2014
The aim of the frequent subgraph mining task is to find frequently occurring subgraphs in a large... more The aim of the frequent subgraph mining task is to find frequently occurring subgraphs in a large graph database. However, this task is a thriving challenge, as graph and subgraph isomorphisms play a key role throughout the computations. Since subgraph isomorphism testing is a hard problem, subgraph miners are exponential in runtime. To alleviate the complexity issue, we propose to introduce a bias in the projection operator and instead of using the costly subgraph isomorphism projection, one can use a polynomial projection having a semantically-valid structural interpretation. This paper presents a new projection operator for graphs named AC-projection, which exhibits nice theoretical complexity properties. We study the size of the search space as well as some practical properties of the projection operator. We also introduce a novel breadth-first algorithm for frequent AC-reduced subgraphs mining. Then, we prove experimentally that we can achieve an important performance gain (pol...
Extraction et Gestion des Connaissances, 2011
The Concept lattice and their Applications conference series began as a response to the increasin... more The Concept lattice and their Applications conference series began as a response to the increasing use of concept lattices in practical applications. These proceedings reflect the diversity of interest in concept lattices by the scientific community with papers ranging from the analysis of algorithms in data mining to teaching mathematics. CLA is structured as a triple peer-review process: conference paper, presentation and extended journal paper. 48 papers were submitted to CLA2007 and 24 papers were selected for presentation. All papers were peer reviewed by at least 3 members of the program committee with the final decision on acceptance determined from the reviews by the program chairs. Based on the presentation of the final papers, outstanding contributions from this volume will be invited to submit an extended version of their conference paper for a special volume on concept lattice applications for the International Journal of General Systems. These papers will again be peer-...
The paper presents a new projection operator for graphs named AC-projection, which exhibits nice ... more The paper presents a new projection operator for graphs named AC-projection, which exhibits nice theoretical complexity properties unlike to the graph isomorphism operator typically used in graph mining. We study the size of the search space as well as some practical properties of the projection operator. We also introduce a novel breadth-first algorithm for frequent AC-reduced subgraphs mining. Then, we prove experimentally that we can achieve an important performance gain (polynomial complexity projection) without or with non-significant loss of discovered patterns in terms of quality.
The generalization of policies in reinforcement learning is a main issue, both from the theoretic... more The generalization of policies in reinforcement learning is a main issue, both from the theoretical model point of view and for their applicability. However, generalizing from a set of examples or searching for regularities is a problem which has already been intensively studied in machine learning. Our work uses techniques in which generalizations are constrained by a language bias, in order to regroup similar states. Such generalizations are principally based on the properties of concept lattices. To guide the possible groupings of similar environment's states, we propose a general algebraic framework, considering the generalization of policies through a set partition of the states and using a language bias as an a priori knowledge. We give an application as an example of our approach by proposing and experimenting a bottom-up algorithm.
Uploads
Papers by Michel Liquiere