HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Theses

Une approche stochastique à base d’arbres aléatoires pour le calcul de dissimilarités : application au clustering pour diverses structures de données

Kevin Dalleau 1
1 ORPAILLEUR - Knowledge representation, reasonning
Inria Nancy - Grand Est, LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : The notion of distance, and more generally of dissimilarity, is an important one in data mining, especially in unsupervised approaches. The algorithms belonging to this class of methods aim at grouping objects in an homogeneous way, and many of them rely on a notion of dissimilarity, in order to quantify the proximity between objects. The choice of algorithms as well as that of dissimilarities is not trivial. Several elements can motivate these choices, such as the type of data – homogeneous data or not –, their representation – feature vectors, graphs –, or some of their characteristics – highly correlated, noisy, etc. –. Although many measures exist, their choice can become complex in some specific settings. This leads to additional complexity in data mining tasks. In this thesis, we present a new approach for computing dissimilarities based on random trees. It is an original approach, which has several advantages such as a great versatility. Indeed, using different dissimilarity calculation modules that we can plug to the method, it becomes possible to apply it in various settings. In particular, we present in this document two modules, enabling the computation of dissimilarities - and, in fine, clustering - on data structured as feature vectors, and on data in the form of graphs. We discuss the very promising results obtained by this approach, as well as the numerous perspectives that it opens, such as the computation of dissimilarity in the framework of attributed graphs, through a unified approach.
Complete list of metadata

https://hal.univ-lorraine.fr/tel-03598291
Contributor : Thèses Ul Connect in order to contact the contributor
Submitted on : Friday, March 4, 2022 - 5:56:46 PM
Last modification on : Saturday, March 5, 2022 - 3:30:51 AM

File

DDOC_T_2021_0181_DALLEAU.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-03598291, version 1

Citation

Kevin Dalleau. Une approche stochastique à base d’arbres aléatoires pour le calcul de dissimilarités : application au clustering pour diverses structures de données. Informatique [cs]. Université de Lorraine, 2021. Français. ⟨NNT : 2021LORR0181⟩. ⟨tel-03598291⟩

Share

Metrics

Record views

82

Files downloads

12