Skip to Main content Skip to Navigation
Theses

Carte topologique pour données qualitatives: application à la reconnaissance automatique de la densité du trafic routier

Abstract : This research concerns the qualitative data processing with neuronal methods. Various not supervised models of training are proposed. This thesis was done In the research departement of RENAULT. The Work was focused on the development of traffic recognition model.

The first model suggested in this thesis is dedicated to binary data 'BTM'. It is a model of vectorial quantization of topological map. This model takes the principal characteristics of the topological map. For this new model, in order to take into account specificities of the binary data, we chose to change the usual metric used in the topological map by the Hamming distance.

The second model is the first probabilistic model of topological map dedicated to the binary data. This model is inspired from work which models a distribution by a mixture of mixture of Bernoulli laws.

The third model is a new model of topological map called CTM (Categorical topological Map) adapted to the not supervised classification of the multidimensional qualitative data. This new model preserves the principal characteristics of the two preceding models. In order to maximize the likelihood, CTM uses in a traditional way algorithm EM.

In this memory, we introduce the applicability for the work undertaken at RENAULT. We will detail the original contribution of our work: use of qualitative information to treat the traffic recognition. We will expose the various analyses done with the application of the algorithms proposed.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00161698
Contributor : Carlos Mejia <>
Submitted on : Wednesday, July 11, 2007 - 2:17:51 PM
Last modification on : Tuesday, December 8, 2020 - 3:42:18 AM
Long-term archiving on: : Thursday, April 8, 2010 - 11:00:59 PM

Identifiers

  • HAL Id : tel-00161698, version 1

Citation

Mustapha Lebbah. Carte topologique pour données qualitatives: application à la reconnaissance automatique de la densité du trafic routier. Mathématiques [math]. Université de Versailles-Saint Quentin en Yvelines, 2003. Français. ⟨tel-00161698⟩

Share

Metrics

Record views

783

Files downloads

712