Overlapping Mixtures of Gaussian Processes for the Data Association Problem

[edit]

Miguel Lázaro Gredilla
Steven Van Vaerenbergh
Neil D. Lawrence, University of Sheffield

Pattern Recognition 10

Related Material

Abstract

In this work we introduce a mixture of GPs to address the data association problem, i.e., to label a group of observations according to the sources that generated them. Unlike several previously proposed GP mixtures, the novel mixture has the distinct characteristic of using no gating function to determine the association of samples and mixture components. Instead, all the GPs in the mixture are global and samples are clustered following “trajectories” across input space. We use a non-standard variational Bayesian algorithm to efficiently recover sample labels and learn the hyperparameters. We show how multi-object tracking problems can be disambiguated and also explore the characteristics of the model in traditional regression settings.


@Article{lazaro-overlapping11,
  title = 	 {Overlapping Mixtures of Gaussian Processes for the Data Association Problem},
  journal =  	 {Pattern Recognition},
  author = 	 {Miguel Lázaro Gredilla and Steven Van Vaerenbergh and Neil D. Lawrence},
  year = 	 {2011},
  volume = 	 {10},
  number =       {4},
  month = 	 {00},
  edit = 	 {https://github.com/lawrennd//publications/edit/gh-pages/_posts/2011-01-01-lazaro-overlapping11.md},
  url =  	 {http://inverseprobability.com/publications/lazaro-overlapping11.html},
  abstract = 	 {In this work we introduce a mixture of GPs to address the data association problem, i.e., to label a group of observations according to the sources that generated them. Unlike several previously proposed GP mixtures, the novel mixture has the distinct characteristic of using no gating function to determine the association of samples and mixture components. Instead, all the GPs in the mixture are global and samples are clustered following “trajectories” across input space. We use a non-standard variational Bayesian algorithm to efficiently recover sample labels and learn the hyperparameters. We show how multi-object tracking problems can be disambiguated and also explore the characteristics of the model in traditional regression settings.},
  key = 	 {Lazaro-overlapping11},
  doi = 	 {10.1016/j.patcog.2011.10.004},
  OPTgroup = 	 {}
 

}
%T Overlapping Mixtures of Gaussian Processes for the Data Association Problem
%A Miguel Lázaro Gredilla and Steven Van Vaerenbergh and Neil D. Lawrence
%B 
%C Pattern Recognition
%D 
%F lazaro-overlapping11
%J Pattern Recognition	
%P --
%R 10.1016/j.patcog.2011.10.004
%U http://inverseprobability.com/publications/lazaro-overlapping11.html
%V 10
%N 4
%X In this work we introduce a mixture of GPs to address the data association problem, i.e., to label a group of observations according to the sources that generated them. Unlike several previously proposed GP mixtures, the novel mixture has the distinct characteristic of using no gating function to determine the association of samples and mixture components. Instead, all the GPs in the mixture are global and samples are clustered following “trajectories” across input space. We use a non-standard variational Bayesian algorithm to efficiently recover sample labels and learn the hyperparameters. We show how multi-object tracking problems can be disambiguated and also explore the characteristics of the model in traditional regression settings.
TY  - CPAPER
TI  - Overlapping Mixtures of Gaussian Processes for the Data Association Problem
AU  - Miguel Lázaro Gredilla
AU  - Steven Van Vaerenbergh
AU  - Neil D. Lawrence
PY  - 2011/01/01
DA  - 2011/01/01	
ID  - lazaro-overlapping11	
SP  - 
EP  - 
DO  - 10.1016/j.patcog.2011.10.004
UR  - http://inverseprobability.com/publications/lazaro-overlapping11.html
AB  - In this work we introduce a mixture of GPs to address the data association problem, i.e., to label a group of observations according to the sources that generated them. Unlike several previously proposed GP mixtures, the novel mixture has the distinct characteristic of using no gating function to determine the association of samples and mixture components. Instead, all the GPs in the mixture are global and samples are clustered following “trajectories” across input space. We use a non-standard variational Bayesian algorithm to efficiently recover sample labels and learn the hyperparameters. We show how multi-object tracking problems can be disambiguated and also explore the characteristics of the model in traditional regression settings.
ER  -

Lázaro Gredilla, M., Van Vaerenbergh, S. & Lawrence, N.D.. (2011). Overlapping Mixtures of Gaussian Processes for the Data Association Problem. Pattern Recognition 10(4):-