Skip to Main content Skip to Navigation
Conference papers

Recursive algorithm for estimation of mixture models on Riemannian symmetric space

Abstract : In many fields such as medical imaging, computer vision and radar signal processing, we are lead to study mixtures of distributions in the Riemannian symmetric space. This paper proposes a recursive algorithm for estimating parameters and simultaneously selecting the number of components of a mixture model on Riemannian symmetric space. The idea is to initialize the estimation process from a large number of components K0 and introduce a prior distribution of the membership weights to express our preference for compact models. Using the Rao-Fisher information gradient to update the parameters, in each iteration the prior drives the irrelevant components to extinction. This algorithm could be applied to estimate the models on symmetric space. This algorithm is simple to use. Indeed, it is robust with respect to the choice of initial values. Moreover, it can select the number of components automatically. We illustrate this paper by some experiments.
Complete list of metadata

Cited literature [16 references]  Display  Hide  Download
Contributor : Jialun Zhou Connect in order to contact the contributor
Submitted on : Monday, September 7, 2020 - 10:49:31 AM
Last modification on : Wednesday, November 3, 2021 - 5:08:03 AM
Long-term archiving on: : Wednesday, December 2, 2020 - 2:53:21 PM


Files produced by the author(s)


  • HAL Id : hal-02927638, version 1


Jialun Zhou, Nicolas Le Bihan, Salem Said. Recursive algorithm for estimation of mixture models on Riemannian symmetric space. GRETSI 2019 - XXVIIème Colloque francophone de traitement du signal et des images, Aug 2019, Lille, France. ⟨hal-02927638⟩



Record views


Files downloads