Ref HAL: hal-00698491_v1
DOI: 10.1007/978-3-642-25838-1

Ref HAL: hal-00686027_v2
Résumé: In classical databases, query performance is casually achieved through physical data structures such as caches, indexes and materialized views. In this context, many cost models help select a "best set'' of such data structures. However, this selection task becomes more complex in the cloud. The criterion to optimize is indeed at least two-dimensional, with the monetary cost of using the cloud balancing query response time. Thus, we define in this paper new cost models that fit into the pay-as-you-go paradigm of cloud computing. These cost models help achieve a multi-criteria optimization of the view materialization vs. CPU power consumption problem, under budget constraints. Finally, we present experimental results that provide a first validation of our contribution and show that cloud view materialization is always desirable.

---------

Résumé: For a distributive lattice L, we consider the problem of interpolating functions f : D → L defined on a finite set D ⊆ Ln , by means of lattice polynomial functions of L. Two instances of this problem have already been solved. In the case when L is a distributive lattice with least and greatest elements 0 and 1, Goodstein proved that a function f : {0, 1}n → L can be interpolated by a lattice polynomial function p : Ln → L if an only if f is monotone; in this case, the interpolating polynomial p was shown to be unique. The interpolation problem was also considered in the more general set- ting where L is a distributive lattice, not necessarily bounded, and where D ⊆ Ln is allowed to range over cuboids D = {a1 , b1 } × · · · × {an , bn } with ai , bi ∈ L and ai < bi . In this case, the class of such partial functions that can be interpolated by lattice polynomial functions was completely described. In this paper, we extend these results by completely characterizing the class of lattice functions that can be interpolated by polynomial functions on arbitrary finite subsets D ⊆ Ln . As in the latter setting, interpolating polynomials are not necessarily unique. We provide explicit descriptions of all possible lattice polynomial functions that interpolate these lattice functions, when such an interpolation is available.

Résumé: Integrals are currently used in multiple criteria analysis for synthesizing into a global evaluation the advantages possessed by a po- tential choice. As such, integrals are operators that increase with the criteria evaluations. However, an item may be also evaluated in terms of its defects. Then the more and the greater the defects, the smaller the evaluation should be. An operator that can provide a synthesis of the defects of an item in this sense is called a desintegral. Desintegrals are maximal when no defects at all are present, while integrals are maximal when all advantages are sufficiently present. So, the greater the value of an integral, or a desintegral, the better the corresponding item since advantages are greater, or defects are smaller respectively. Desintegrals implicitly refer to a negative scale, since an order-reversing mapping of the scale used for evaluating each criterion transforms the degree to which the value is advantageous into a degree to which it is disadvanta- geous, and conversely. In this paper, we provide an organised description of counterparts to Sugeno integrals that synthesize positive or negative evaluations in the qualitative framework of a totally ordered residuated lattice equipped with an involutive negation. We exploit three kinds of criteria weighting schemes that are allowed by this algebraic structure.

Ref HAL: hal-00691929_v1
Résumé: CPTS-integrals (Cumulative Prospect Theory Sugeno inte- gral) are agregation functions that return a global rating when the eval- uations are given on a bipolar scale. This paper presents the problem of the elicitation of CPTS-integrals agreeing with a data set composed of pairs concerning some criteria and a global evaluation. Moreover the set of fuzzy measures which are solutions of this inverse problem is identi- fied. When the elicitation of one CPTS-integral is not possible, a set of family of CPTS-integrals is proposed.

Ref HAL: hal-00713095_v1
Résumé: L’accès à une information pertinente, adaptée aux besoins et au contexte de l’utilisateur, est un véritable défi pour la communauté de la recherche d’information (RI). Les systèmes de recherche d’information (SRI) classiques ne prennent pas en compte les facteurs contextuels dans le processus d’appariement et/ou dans la phase de classement des résultats. Ils s’intéressent plutôt aux métriques de pondération et aux formules de calcul de similarité entre les termes de la requête utilisateur et ceux des documents du corpus. Dans cet article, nous proposons une approche de RI qui prend en compte à la fois les paramètres contextuels et les centres d’intérêts de l’utilisateur obtenus par un processus de segmentation thématique des documents. Ainsi, nous proposons une adaptation de la formule TF-IDF pour considérer les unités thématiques, représentées sous forme de fragments de texte, lors de la phase de représentation et de classification des documents . Nos expérimentations réalisées sur le corpus 20 Newsgroups montrent une nette amélioration des taux de précision par rapport aux SRI classiques qui ne considèrent pas les facteurs contextuels.

Ref HAL: halshs-00677401_v1
Résumé: Data warehouses and Online Analysis Processing (OLAP) have acknowledged and efficient solutions for helping in the decision-making process. Through OLAP operators, online analysis enables the decision-maker to navigate and view data represented in a multidimensional manner. But when the data or objects to be analyzed are complex, it is necessary to redefine and enhance the abilities of the OLAP. In this paper, we suggest combining OLAP and data mining in order to create a new visualization operator for complex data or objects. This operator uses the correspondence analysis method and we call it VOCoDa (Visualization Operator for Complex Data).