Vignette d'image

Jabbari, Simin

Nom
Jabbari, Simin
Affiliation principale
Fonction
Ancien.ne collaborateur.trice
Identifiants
https://libra.unine.ch/handle/123456789/1005

Résultat de la recherche

Filtres

2
2
2
Réinitialiser les filtres

Paramètres

Voici les éléments 1 - 2 sur 2
  • Publication
    Accès libre
    Parallel Execution of Binary-Based NextClosure Algorithm
    (2016-7-18)
    Jabbari, Simin 
    Formal concept analysis (FCA) has become a popular method for analyzing data across various domains in which data bases can be analyzed regardless of their contexts. With its properties FCA is of big interest in the context of Big Data. However, the complexity of the basic FCA analysis algorithms often prohibits its use in general production tool chains for data analysis. In this paper we show how to overcome some of these problems. In the first step we show how to implement the well known NextClosure in efficient way in Python (a preferred language in the context of ad-hoc data analysis) which is several times faster the other published algorithms. In the second step we show how our implementation can be parallelized on common hardware by strictly using the best sequential algorithm which di↵ers in an important way form so far published parallel algorithms for FCA.
  • Publication
    Accès libre
    A Hybrid Algorithm for Generating Formal Concepts and Building Concept Lattice Using NextClosure and Nourine Algorithms
    (2016-7-18)
    Jabbari, Simin 
    Concept lattice produced from a set of formal concepts is used for representing concept hierarchy and has many applications in knowledge representation and data mining. Different algorithms have been proposed in the past for efficiently generating formal concepts and building concept lattices. In this paper we introduce the idea of combining existing algorithms in FCA with the aim of benefiting from their specific advantages. As an example, we propose a hybrid model that utilizes the NextClosure (NC) algorithm for generating formal concepts and parts of the Nourine algorithm for building concept lattices.We compare the proposed hybrid model with two of its counterparts: pure NC and pure Nourine. Our experiments show that the hybrid model always outperforms pure NC and for very large datasets can surpass the pure Nourine, as well.