Gaussian process models for periodicity detection

Abstract : We consider the problem of detecting and quantifying the periodic component of a function given noise-corrupted observations of a limited number of input/output tuples. Our approach is based on Gaussian process regression which provides a flexible non-parametric framework for modelling periodic data. We introduce a novel decomposition of the covariance function as the sum of periodic and aperiodic kernels. This decomposition allows for the creation of sub-models which capture the periodic nature of the signal and its complement. To quantify the periodicity of the signal, we derive a periodicity ratio which reflects the uncertainty in the fitted sub-models. Although the method can be applied to many kernels, we give a special emphasis to the Mat\'ern family, from the expression of the reproducing kernel Hilbert space inner product to the implementation of the associated periodic kernels in a Gaussian process toolkit. The proposed method is illustrated by considering the detection of periodically expressed genes in the arabidopsis genome.
Type de document :
Article dans une revue
PeerJ Computer Science, PeerJ, 2016
Liste complète des métadonnées
Contributeur : Nicolas Durrande <>
Soumis le : jeudi 18 août 2016 - 15:54:38
Dernière modification le : mardi 23 octobre 2018 - 14:36:08


Fichiers produits par l'(les) auteur(s)


  • HAL Id : hal-00805468, version 3
  • ARXIV : 1303.7090


Nicolas Durrande, James Hensman, Magnus Rattray, Neil D. Lawrence. Gaussian process models for periodicity detection. PeerJ Computer Science, PeerJ, 2016. 〈hal-00805468v3〉



Consultations de la notice


Téléchargements de fichiers