Time Series Filtering, Smoothing and Learning using the Kernel Kalman Filter

Liva Ralaivola; Florence D'Alché-Buc

HAL: hal-00341953, version 1

DOI: 10.1109/IJCNN.2005.1556088

Detailed view

Export this paper

Proceedings of IEEE International Joint Conference on Neural Networks 2005 (IJCNN 2005), Canada (2005)

Time Series Filtering, Smoothing and Learning using the Kernel Kalman Filter

Liva Ralaivola ¹, Florence D'Alché-Buc ²

(2005)

In this paper, we propose a new model, the kernel Kalman Filter, to perform various nonlinear time series processing. This model is based on the use of Mercer kernel functions in the framework of the Kalman filter or linear dynamical systems. Thanks to the kernel trick, all the equations involved in our model to perform filtering, smoothing and learning tasks, only require matrix algebra calculus whilst providing the ability to model complex time series. In particular, it is possible to learn dynamics from some nonlinear noisy time series implementing an exact expectation-maximization procedure.

1:	Laboratoire d'informatique Fondamentale de Marseille (LIF)

	CNRS : UMR6166 – Université de la Méditerranée - Aix-Marseille II – Université de Provence - Aix-Marseille I

2:	Informatique, Biologie Intégrative et Systèmes Complexes (IBISC)

	CNRS : FRE3190 – Université d'Evry-Val d'Essonne

Subject

Cognitive science/Computer science

Computer Science/Bioinformatics

Life Sciences/Quantitative Methods

Computer Science/Modeling and Simulation

Computer Science/Learning

Computer Science/Databases


hal-00341953, version 1
http://hal.archives-ouvertes.fr/hal-00341953
oai:hal.archives-ouvertes.fr:hal-00341953
From: Frédéric Davesne <>
Submitted on: Wednesday, 26 November 2008 14:09:57
Updated on: Wednesday, 11 February 2009 19:54:08