Catálogo de publicaciones - libros

Data mining, with its objective to efficiently discover valuable and inherent information from large databases, is particularly sensitive to misuse. Therefore an interesting new direction for data mining research is the development of techniques that incorporate privacy concerns and to develop accurate models without access to precise information in individual data records. The difficulty lies in the fact that the two metrics for evaluating privacy preserving data mining methods: privacy and accuracy are typically contradictory in nature. We address privacy preserving mining on distributed data in this paper and present an algorithm, based on the combination of probabilistic approach and cryptographic approach, to protect high privacy of individual information and at the same time acquire a high level of accuracy in the mining result.

In an emergency situation, the actors need an assistance allowing them to react swiftly and efficiently. In this prospect, we present in this paper a decision support system that aims to prepare actors in a crisis situation thanks to a decision-making support. The global architecture of this system is presented in the first part. Then we focus on a part of this system which is designed to represent the information of the current situation. This part is composed of a multiagent system that is made of factual agents. Each agent carries a semantic feature and aims to represent a partial part of a situation. The agents develop thanks to their interactions by comparing their semantic features using proximity measures and according to specific ontologies.

Negotiation is a crucial stage of Web Services interaction lifecycle. By exchanging a sequence of proposals in the negotiation stage, a service provider and a consumer try to establish a formal contract, to specify agreed terms on the service, particularly terms on non-functional aspects. To react to an ever-changing environment, flexible negotiation strategies that can make adjustable rates of concession should be adopted. This paper presents such flexible strategies for Web Services negotiation. In a negotiation round, the negotiation strategies first examine the environment situations by evaluating certain factors, which include time, resources, number of counterparts and current proposals from the counterparts. For each factor, there is a corresponding function that suggests the amount of concession in terms of that factor. Then considering the importance of each service attribute, the target concession per attribute is determined. As a final step, a set of experimental tests is executed to evaluate the performance of the negotiation strategies.

Classifier ensembles is an active area of research within the machine learning community. One of the most successful techniques is , where an algorithm (typically a decision tree inducer) is applied over several different training sets, obtained applying sampling with replacement to the original database. In this paper we define a framework where sampling with and without replacement can be viewed as the extreme cases of a more general process, and analyze the performance of the extension of bagging to such framework.

The multiplication of execution contexts for multimedia documents requires the adaptation of the document specification to the particularities of the contexts. In this paper, we propose to apply a semantic approach for multimedia document adaptation to the spatio-temporal dimension of documents. To guarantee that the adapted document is close to the initial one respecting adaptation constraints, we define proximities for adapting static documents ( documents without animations) and animated documents. Moreover, we show that these proximities can be refined according to multimedia object properties ( images, videos...). The approach is illustrated by an example.

In daily life, description and location of certain objects on the web are much dependent on color names. Therefore, a maturely-implemented matching subsystem for color descriptions will certainly facilitate web applications in the domains concerned, such as image retrieval, clothing search, etc. However, both keyword matching and semantic mediation by the current ontologies are confronted with difficulties in precisely evaluating the similarity between color descriptions, which requests the exploitation of “deeper” semantics to bridge the semantic gap. What with the inherent variability and imprecision characterizing color naming, this paper proposes a novel approach for defining (1) the fuzzy semantics of color names on the HSL color space, and (2) the associated measures to evaluate the similarity between two fuzzified color descriptions. The experimental results have preliminarily shown the strength of the deeper semantics surpassing the ability of both keywords and WordNet, in dealing with the matching problem of color descriptions.

Managing emails from list-servs is an open problem that we believe may be partially resolved by the introduction of a principled, argumentation-based approach towards their representation. We propose an argumentation ontology, called “Argontonaut,” which has been developed for the domain of standards-driven W3C mailing lists. We use the extensible nature of RDF to fuse an argumentation-based approach with one grounded in an issue-management system. We motivate our ontology with reference to the domain and propose future work in this area.

Clustering techniques are considered as efficient tools for partitioning data sets in order to get homogeneous clusters of objects. However, the reality is connected to uncertainty by nature, and these standard algorithms of clustering do not deal with this uncertainty pervaded in their parameters. In this paper we develop a clustering method in an uncertain context based on the K-modes method and the belief function theory. This so-called belief K-modes method (BKM) provides a new clustering technique handling uncertainty in the attribute values of objects in both the clusters’ construction task and the classification one.

Although speech and language processing techniques achieved a relative maturity during the last decade, designing a spoken dialogue system is still a tailoring task because of the great variability of factors to take into account. Rapid design and reusability across tasks of previous work is made very difficult. For these reasons, machine learning methods applied to dialogue strategy optimization has become a leading subject of researches since the mid 90’s. In this paper, we describe an experiment of reinforcement learning applied to the optimization of speech-based database querying. We will especially emphasize on the sensibility of the method relatively to the dialogue modeling parameters in the framework of the Markov decision processes, namely the state space and the reinforcement signal. The evolution of the design will be exposed as well as results obtained on a simple real application.

We consider the problem of exploring an unknown environment with an intelligent virtual agent. Traditionally research efforts to address the exploration and mapping problem have focused on the graph-based space representations and the graph search algorithms. In this paper, we propose DFS-RTA* and DFS-PHA*, two real-time graph search algorithms for exploring and mapping an unknown environment. Both algorithms are based upon the simple depth-first search strategy. However, they adopt different real-time shortest path-finding methods for fast backtracking to the last unexhausted node. Through some experiments with a virtual agent deploying in a 3D interactive computer game environment, we confirm completeness and efficiency of two algorithms.