27 juil
27/07/2017 14:00


Soutenance de thèse : Sergio PEIGNIER

Subspace clustering on static datasets and dynamic data streams using bio-inspired algorithms

Doctorant : Sergio PEIGNIER 

Laboratoire INSA : LIRIS
Ecole doctorale : EDA512 : InfoMaths

Subspace clustering, is a recent data mining task that was developed to deal with high dimensional datasets. This task is recognized as more general and complicated than standard clustering, since it aims to detect groups of similar objects and at the same time the subspaces where these similarities appear. The different subspace clustering algorithms that have been proposed in the literature rely on very different algorithmic foundations. Among these approaches, evolutionary algorithms have been under-explored, even if these techniques have proven valuable for other NP- hard problems. The aim of this thesis was to take advantage of new knowledge from evolutionary biology in order to conceive evolutionary subspace clustering algorithms for static datasets and dynamic data streams. Chameleoclust, the first algorithm proposed in this work, takes advantage of the large degree of freedom provided by several bio-like features such as a variable genome length, both functional and non-functional elements and mutation operators including chromosomal rearrangements. KymeroClust, our second algorithm, is a median-based approach that relies on a cornerstone evolutionary mechanism: duplication and divergence of genes. SubMorphoStream, the last one, tackles the subspace clustering task over dynamic data streams. It relies on two major mechanisms that favor fast adaptation of bacteria to changing environments: gene amplification and foreign genetic material uptake. All these algorithms were compared to the main state-of-the- art techniques, obtaining competitive results. This suggests that they are useful complementary data mining tools.

In addition two applications called EvoWave and EvoMove have been developed to assess the capacity of these algorithms to address real world problems. EvoWave and EvoMove handled respectively the analysis of wifi signal contexts and of dancer moves captured using motion sensors. Promising results also advocate for the use of bio-inspired mechanisms to deal with other real world applications.

Informations complémentaires

  • Bâtiment Blaise Pascal, Salle de réunion IF (502.321), 9 avenue Jean Capelle, Villeurbanne.