Fast Nonparametric Clustering of Structured Time-Series

James HensmanMagnus RattrayNeil D. Lawrence
IEEE Transactions on Pattern Analysis and Machine Intelligence, 37:383-393, 2014.

Abstract

In this publication, we combine two Bayesian nonparametric models: the Gaussian Process (GP) and the Dirichlet Process (DP). Our innovation in the GP model is to introduce a variation on the GP prior which enables us to model structured time-series data, i.e. data containing groups where we wish to model inter- and intra-group variability. Our innovation in the DP model is an implementation of a new fast collapsed variational inference procedure which enables us to optimize our variational approximation significantly faster than standard VB approaches. In a biological time series application we show how our model better captures salient features of the data, leading to better consistency with existing biological classifications, while the associated inference algorithm provides a significant speed-up over EM-based variational inference.

Cite this Paper


BibTeX
@Article{Hensman-fast14, title = {Fast Nonparametric Clustering of Structured Time-Series}, author = {Hensman, James and Rattray, Magnus and Lawrence, Neil D.}, journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence}, pages = {383--393}, year = {2014}, volume = {37}, doi = {10.1109/TPAMI.2014.2318711}, pdf = {https://ieeexplore.ieee.org/ielx7/34/7004096/06802369.pdf}, url = {http://inverseprobability.com/publications/hensman-fast14.html}, abstract = {In this publication, we combine two Bayesian nonparametric models: the Gaussian Process (GP) and the Dirichlet Process (DP). Our innovation in the GP model is to introduce a variation on the GP prior which enables us to model structured time-series data, i.e. data containing groups where we wish to model inter- and intra-group variability. Our innovation in the DP model is an implementation of a new fast collapsed variational inference procedure which enables us to optimize our variational approximation significantly faster than standard VB approaches. In a biological time series application we show how our model better captures salient features of the data, leading to better consistency with existing biological classifications, while the associated inference algorithm provides a significant speed-up over EM-based variational inference.} }
Endnote
%0 Journal Article %T Fast Nonparametric Clustering of Structured Time-Series %A James Hensman %A Magnus Rattray %A Neil D. Lawrence %J IEEE Transactions on Pattern Analysis and Machine Intelligence %D 2014 %F Hensman-fast14 %P 383--393 %R 10.1109/TPAMI.2014.2318711 %U http://inverseprobability.com/publications/hensman-fast14.html %V 37 %X In this publication, we combine two Bayesian nonparametric models: the Gaussian Process (GP) and the Dirichlet Process (DP). Our innovation in the GP model is to introduce a variation on the GP prior which enables us to model structured time-series data, i.e. data containing groups where we wish to model inter- and intra-group variability. Our innovation in the DP model is an implementation of a new fast collapsed variational inference procedure which enables us to optimize our variational approximation significantly faster than standard VB approaches. In a biological time series application we show how our model better captures salient features of the data, leading to better consistency with existing biological classifications, while the associated inference algorithm provides a significant speed-up over EM-based variational inference.
RIS
TY - JOUR TI - Fast Nonparametric Clustering of Structured Time-Series AU - James Hensman AU - Magnus Rattray AU - Neil D. Lawrence DA - 2014/01/01 ID - Hensman-fast14 VL - 37 SP - 383 EP - 393 DO - 10.1109/TPAMI.2014.2318711 L1 - https://ieeexplore.ieee.org/ielx7/34/7004096/06802369.pdf UR - http://inverseprobability.com/publications/hensman-fast14.html AB - In this publication, we combine two Bayesian nonparametric models: the Gaussian Process (GP) and the Dirichlet Process (DP). Our innovation in the GP model is to introduce a variation on the GP prior which enables us to model structured time-series data, i.e. data containing groups where we wish to model inter- and intra-group variability. Our innovation in the DP model is an implementation of a new fast collapsed variational inference procedure which enables us to optimize our variational approximation significantly faster than standard VB approaches. In a biological time series application we show how our model better captures salient features of the data, leading to better consistency with existing biological classifications, while the associated inference algorithm provides a significant speed-up over EM-based variational inference. ER -
APA
Hensman, J., Rattray, M. & Lawrence, N.D.. (2014). Fast Nonparametric Clustering of Structured Time-Series. IEEE Transactions on Pattern Analysis and Machine Intelligence 37:383-393 doi:10.1109/TPAMI.2014.2318711 Available from http://inverseprobability.com/publications/hensman-fast14.html.

Related Material