Return to results
Conference Proceedings (2015), 25ème Édition du Colloque GRETSI, École Normale Supérieure de Lyon

Distribution hybride pour la modélisation de données à deux queues lourdes: Application sur les données neuronales

DEBBABI N., KRATZ Marie , MBOUP M., EL ASMI S.

A new hybrid model for two heavy tailed data modelling is proposed in this study. The proposed model is a weighted three-components distribution: a Gaussian distribution, to model the mean behavior of the data, linked to two generalized Pareto distributions, modelling the extreme ones. An unsupervised iterative algorithm is then developed to estimate accurately the junction points between the three distributions, the parameters of these latter as well as the weights of the hybrid model. An application on real extracellular neural recordings isdeveloped to evaluate the performance of the proposed hybrid model, compared to the normal distribution.

DEBBABI, N., KRATZ, M., MBOUP, M. and EL ASMI, S. (2015). Distribution hybride pour la modélisation de données à deux queues lourdes: Application sur les données neuronales. In: 25ème Édition du Colloque GRETSI. École Normale Supérieure de Lyon.