The Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling

Michael Betancourt
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:533-540, 2015.

Abstract

Leveraging the coherent exploration of Hamiltonian flow, Hamiltonian Monte Carlo produces computationally efficient Monte Carlo estimators, even with respect to complex and high-dimensional target distributions. When confronted with data-intensive applications, however, the algorithm may be too expensive to implement, leaving us to consider the utility of approximations such as data subsampling. In this paper I demonstrate how data subsampling fundamentally compromises the scalability of Hamiltonian Monte Carlo.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-betancourt15, title = {The Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling}, author = {Betancourt, Michael}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {533--540}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/betancourt15.pdf}, url = {https://proceedings.mlr.press/v37/betancourt15.html}, abstract = {Leveraging the coherent exploration of Hamiltonian flow, Hamiltonian Monte Carlo produces computationally efficient Monte Carlo estimators, even with respect to complex and high-dimensional target distributions. When confronted with data-intensive applications, however, the algorithm may be too expensive to implement, leaving us to consider the utility of approximations such as data subsampling. In this paper I demonstrate how data subsampling fundamentally compromises the scalability of Hamiltonian Monte Carlo.} }
Endnote
%0 Conference Paper %T The Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling %A Michael Betancourt %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-betancourt15 %I PMLR %P 533--540 %U https://proceedings.mlr.press/v37/betancourt15.html %V 37 %X Leveraging the coherent exploration of Hamiltonian flow, Hamiltonian Monte Carlo produces computationally efficient Monte Carlo estimators, even with respect to complex and high-dimensional target distributions. When confronted with data-intensive applications, however, the algorithm may be too expensive to implement, leaving us to consider the utility of approximations such as data subsampling. In this paper I demonstrate how data subsampling fundamentally compromises the scalability of Hamiltonian Monte Carlo.
RIS
TY - CPAPER TI - The Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling AU - Michael Betancourt BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-betancourt15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 533 EP - 540 L1 - http://proceedings.mlr.press/v37/betancourt15.pdf UR - https://proceedings.mlr.press/v37/betancourt15.html AB - Leveraging the coherent exploration of Hamiltonian flow, Hamiltonian Monte Carlo produces computationally efficient Monte Carlo estimators, even with respect to complex and high-dimensional target distributions. When confronted with data-intensive applications, however, the algorithm may be too expensive to implement, leaving us to consider the utility of approximations such as data subsampling. In this paper I demonstrate how data subsampling fundamentally compromises the scalability of Hamiltonian Monte Carlo. ER -
APA
Betancourt, M.. (2015). The Fundamental Incompatibility of Scalable Hamiltonian Monte Carlo and Naive Data Subsampling. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:533-540 Available from https://proceedings.mlr.press/v37/betancourt15.html.

Related Material