Pitfalls in the use of Parallel Inference for the Dirichlet Process

Yarin Gal, Zoubin Ghahramani
; Proceedings of the 31st International Conference on Machine Learning, PMLR 32(2):208-216, 2014.

Abstract

Recent work done by Lovell, Adams, and Mansingka (2012) and Williamson, Dubey, and Xing (2013) has suggested an alternative parametrisation for the Dirichlet process in order to derive non-approximate parallel MCMC inference for it - work which has been picked-up and implemented in several different fields. In this paper we show that the approach suggested is impractical due to an extremely unbalanced distribution of the data. We characterise the requirements of efficient parallel inference for the Dirichlet process and show that the proposed inference fails most of these requirements (while approximate approaches often satisfy most of them). We present both theoretical and experimental evidence, analysing the load balance for the inference and showing that it is independent of the size of the dataset and the number of nodes available in the parallel implementation. We end with suggestions of alternative paths of research for efficient non-approximate parallel inference for the Dirichlet process.

Cite this Paper


BibTeX
@InProceedings{pmlr-v32-gal14, title = {Pitfalls in the use of Parallel Inference for the Dirichlet Process}, author = {Yarin Gal and Zoubin Ghahramani}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {208--216}, year = {2014}, editor = {Eric P. Xing and Tony Jebara}, volume = {32}, number = {2}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/gal14.pdf}, url = {http://proceedings.mlr.press/v32/gal14.html}, abstract = {Recent work done by Lovell, Adams, and Mansingka (2012) and Williamson, Dubey, and Xing (2013) has suggested an alternative parametrisation for the Dirichlet process in order to derive non-approximate parallel MCMC inference for it - work which has been picked-up and implemented in several different fields. In this paper we show that the approach suggested is impractical due to an extremely unbalanced distribution of the data. We characterise the requirements of efficient parallel inference for the Dirichlet process and show that the proposed inference fails most of these requirements (while approximate approaches often satisfy most of them). We present both theoretical and experimental evidence, analysing the load balance for the inference and showing that it is independent of the size of the dataset and the number of nodes available in the parallel implementation. We end with suggestions of alternative paths of research for efficient non-approximate parallel inference for the Dirichlet process.} }
Endnote
%0 Conference Paper %T Pitfalls in the use of Parallel Inference for the Dirichlet Process %A Yarin Gal %A Zoubin Ghahramani %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-gal14 %I PMLR %J Proceedings of Machine Learning Research %P 208--216 %U http://proceedings.mlr.press %V 32 %N 2 %W PMLR %X Recent work done by Lovell, Adams, and Mansingka (2012) and Williamson, Dubey, and Xing (2013) has suggested an alternative parametrisation for the Dirichlet process in order to derive non-approximate parallel MCMC inference for it - work which has been picked-up and implemented in several different fields. In this paper we show that the approach suggested is impractical due to an extremely unbalanced distribution of the data. We characterise the requirements of efficient parallel inference for the Dirichlet process and show that the proposed inference fails most of these requirements (while approximate approaches often satisfy most of them). We present both theoretical and experimental evidence, analysing the load balance for the inference and showing that it is independent of the size of the dataset and the number of nodes available in the parallel implementation. We end with suggestions of alternative paths of research for efficient non-approximate parallel inference for the Dirichlet process.
RIS
TY - CPAPER TI - Pitfalls in the use of Parallel Inference for the Dirichlet Process AU - Yarin Gal AU - Zoubin Ghahramani BT - Proceedings of the 31st International Conference on Machine Learning PY - 2014/01/27 DA - 2014/01/27 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-gal14 PB - PMLR SP - 208 DP - PMLR EP - 216 L1 - http://proceedings.mlr.press/v32/gal14.pdf UR - http://proceedings.mlr.press/v32/gal14.html AB - Recent work done by Lovell, Adams, and Mansingka (2012) and Williamson, Dubey, and Xing (2013) has suggested an alternative parametrisation for the Dirichlet process in order to derive non-approximate parallel MCMC inference for it - work which has been picked-up and implemented in several different fields. In this paper we show that the approach suggested is impractical due to an extremely unbalanced distribution of the data. We characterise the requirements of efficient parallel inference for the Dirichlet process and show that the proposed inference fails most of these requirements (while approximate approaches often satisfy most of them). We present both theoretical and experimental evidence, analysing the load balance for the inference and showing that it is independent of the size of the dataset and the number of nodes available in the parallel implementation. We end with suggestions of alternative paths of research for efficient non-approximate parallel inference for the Dirichlet process. ER -
APA
Gal, Y. & Ghahramani, Z.. (2014). Pitfalls in the use of Parallel Inference for the Dirichlet Process. Proceedings of the 31st International Conference on Machine Learning, in PMLR 32(2):208-216

Related Material