Multivariate online kernel density estimation with Gaussian kernels

We propose a novel approach to online estimation of probability density functions, which is based on kernel density estimation (KDE). The method maintains and updates a non-parametric model of the observed data, from which the KDE can be calculated. We propose an online bandwidth estimation approach...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Pattern recognition Jg. 44; H. 10; S. 2630 - 2642
Hauptverfasser: Kristan, Matej, Leonardis, Aleš, Skočaj, Danijel
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Kidlington Elsevier Ltd 01.10.2011
Elsevier
Schlagworte:
Applied sciences
Artificial intelligence
Complexity
Compressing
Computer science; control theory; systems
Density
Exact sciences and technology
Gaussian
Gaussian mixture models
Information, signal and communications theory
Kernel density estimation
Kernels
On-line systems
Online
Online models
Pattern recognition
Pattern recognition. Digital image processing. Computational geometry
Probability density estimation
Signal processing
State of the art
Telecommunications and information theory
Gaussian mixture models
Probability density estimation
Kernel density estimation
Online models
Performance evaluation
Density estimation
State of the art
Non parametric method
Mixture theory
Updating
Pattern recognition
Kernel method
Statistical method
Probability density
Probabilistic assessment
Gaussian process
On line processing
Probability density function
ISSN:0031-3203, 1873-5142
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:We propose a novel approach to online estimation of probability density functions, which is based on kernel density estimation (KDE). The method maintains and updates a non-parametric model of the observed data, from which the KDE can be calculated. We propose an online bandwidth estimation approach and a compression/revitalization scheme which maintains the KDE's complexity low. We compare the proposed online KDE to the state-of-the-art approaches on examples of estimating stationary and non-stationary distributions, and on examples of classification. The results show that the online KDE outperforms or achieves a comparable performance to the state-of-the-art and produces models with a significantly lower complexity while allowing online adaptation. [Display omitted] ► We propose a solution for online estimation of probability density functions. ► We extend the batch kernel density estimators (KDE) to online KDEs (oKDE). ► oKDE's complexity scales sublinearly with the number of samples. ► oKDE outperforms batch KDEs in non-stationary distribution estimation. ► oKDE achieves comparable classification performance to a batch SVM.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0031-3203
1873-5142
DOI:10.1016/j.patcog.2011.03.019