Slice Sampling on Hamiltonian Trajectories

Benjamin Bloem-Reddy, John Cunningham
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:3050-3058, 2016.

Abstract

Hamiltonian Monte Carlo and slice sampling are amongst the most widely used and studied classes of Markov Chain Monte Carlo samplers. We connect these two methods and present Hamiltonian slice sampling, which allows slice sampling to be carried out along Hamiltonian trajectories, or transformations thereof. Hamiltonian slice sampling clarifies a class of model priors that induce closed-form slice samplers. More pragmatically, inheriting properties of slice samplers, it offers advantages over Hamiltonian Monte Carlo, in that it has fewer tunable hyperparameters and does not require gradient information. We demonstrate the utility of Hamiltonian slice sampling out of the box on problems ranging from Gaussian process regression to Pitman-Yor based mixture models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v48-bloem-reddy16, title = {Slice Sampling on Hamiltonian Trajectories}, author = {Bloem-Reddy, Benjamin and Cunningham, John}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {3050--3058}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/bloem-reddy16.pdf}, url = {https://proceedings.mlr.press/v48/bloem-reddy16.html}, abstract = {Hamiltonian Monte Carlo and slice sampling are amongst the most widely used and studied classes of Markov Chain Monte Carlo samplers. We connect these two methods and present Hamiltonian slice sampling, which allows slice sampling to be carried out along Hamiltonian trajectories, or transformations thereof. Hamiltonian slice sampling clarifies a class of model priors that induce closed-form slice samplers. More pragmatically, inheriting properties of slice samplers, it offers advantages over Hamiltonian Monte Carlo, in that it has fewer tunable hyperparameters and does not require gradient information. We demonstrate the utility of Hamiltonian slice sampling out of the box on problems ranging from Gaussian process regression to Pitman-Yor based mixture models.} }
Endnote
%0 Conference Paper %T Slice Sampling on Hamiltonian Trajectories %A Benjamin Bloem-Reddy %A John Cunningham %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-bloem-reddy16 %I PMLR %P 3050--3058 %U https://proceedings.mlr.press/v48/bloem-reddy16.html %V 48 %X Hamiltonian Monte Carlo and slice sampling are amongst the most widely used and studied classes of Markov Chain Monte Carlo samplers. We connect these two methods and present Hamiltonian slice sampling, which allows slice sampling to be carried out along Hamiltonian trajectories, or transformations thereof. Hamiltonian slice sampling clarifies a class of model priors that induce closed-form slice samplers. More pragmatically, inheriting properties of slice samplers, it offers advantages over Hamiltonian Monte Carlo, in that it has fewer tunable hyperparameters and does not require gradient information. We demonstrate the utility of Hamiltonian slice sampling out of the box on problems ranging from Gaussian process regression to Pitman-Yor based mixture models.
RIS
TY - CPAPER TI - Slice Sampling on Hamiltonian Trajectories AU - Benjamin Bloem-Reddy AU - John Cunningham BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-bloem-reddy16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 3050 EP - 3058 L1 - http://proceedings.mlr.press/v48/bloem-reddy16.pdf UR - https://proceedings.mlr.press/v48/bloem-reddy16.html AB - Hamiltonian Monte Carlo and slice sampling are amongst the most widely used and studied classes of Markov Chain Monte Carlo samplers. We connect these two methods and present Hamiltonian slice sampling, which allows slice sampling to be carried out along Hamiltonian trajectories, or transformations thereof. Hamiltonian slice sampling clarifies a class of model priors that induce closed-form slice samplers. More pragmatically, inheriting properties of slice samplers, it offers advantages over Hamiltonian Monte Carlo, in that it has fewer tunable hyperparameters and does not require gradient information. We demonstrate the utility of Hamiltonian slice sampling out of the box on problems ranging from Gaussian process regression to Pitman-Yor based mixture models. ER -
APA
Bloem-Reddy, B. & Cunningham, J.. (2016). Slice Sampling on Hamiltonian Trajectories. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:3050-3058 Available from https://proceedings.mlr.press/v48/bloem-reddy16.html.

Related Material