A Kernel Multiple Change-point Algorithm via Model Selection - Archive ouverte HAL Access content directly
Journal Articles Journal of Machine Learning Research Year : 2019

A Kernel Multiple Change-point Algorithm via Model Selection

Sylvain Arlot
Alain Celisse

Abstract

We tackle the change-point problem with data belonging to a general set. We build a penalty for choosing the number of change-points in the kernel-based method of Harchaoui and Cappé (2007). This penalty generalizes the one proposed by Lebarbier (2005) for one-dimensional signals. We prove a non-asymptotic oracle inequality for the proposed method, thanks to a new concentration result for some function of Hilbert-space valued random variables. Experiments on synthetic data illustrate the accuracy of our method, showing that it can detect changes in the whole distribution of data, even when the mean and variance are constant.
Fichier principal
Vignette du fichier
kernelchpt_hal_v3.pdf (1.16 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-00671174 , version 1 (17-02-2012)
hal-00671174 , version 2 (18-03-2016)
hal-00671174 , version 3 (14-03-2019)

Identifiers

Cite

Sylvain Arlot, Alain Celisse, Zaid Harchaoui. A Kernel Multiple Change-point Algorithm via Model Selection. Journal of Machine Learning Research, 2019, 20 (162), pp.1--56. ⟨hal-00671174v3⟩
1379 View
1071 Download

Altmetric

Share

Gmail Facebook X LinkedIn More