Asynchronous Batch Bayesian Optimisation with Improved Local Penalisation

Ahsan Alvi, Binxin Ru, Jan-Peter Calliess, Stephen Roberts, Michael A. Osborne
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:253-262, 2019.

Abstract

Batch Bayesian optimisation (BO) has been successfully applied to hyperparameter tuning using parallel computing, but it is wasteful of resources: workers that complete jobs ahead of others are left idle. We address this problem by developing an approach, Penalising Locally for Asynchronous Bayesian Optimisation on K Workers (PLAyBOOK), for asynchronous parallel BO. We demonstrate empirically the efficacy of PLAyBOOK and its variants on synthetic tasks and a real-world problem. We undertake a comparison between synchronous and asynchronous BO, and show that asynchronous BO often outperforms synchronous batch BO in both wall-clock time and sample efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-alvi19a, title = {Asynchronous Batch {B}ayesian Optimisation with Improved Local Penalisation}, author = {Alvi, Ahsan and Ru, Binxin and Calliess, Jan-Peter and Roberts, Stephen and Osborne, Michael A.}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {253--262}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/alvi19a/alvi19a.pdf}, url = {https://proceedings.mlr.press/v97/alvi19a.html}, abstract = {Batch Bayesian optimisation (BO) has been successfully applied to hyperparameter tuning using parallel computing, but it is wasteful of resources: workers that complete jobs ahead of others are left idle. We address this problem by developing an approach, Penalising Locally for Asynchronous Bayesian Optimisation on K Workers (PLAyBOOK), for asynchronous parallel BO. We demonstrate empirically the efficacy of PLAyBOOK and its variants on synthetic tasks and a real-world problem. We undertake a comparison between synchronous and asynchronous BO, and show that asynchronous BO often outperforms synchronous batch BO in both wall-clock time and sample efficiency.} }
Endnote
%0 Conference Paper %T Asynchronous Batch Bayesian Optimisation with Improved Local Penalisation %A Ahsan Alvi %A Binxin Ru %A Jan-Peter Calliess %A Stephen Roberts %A Michael A. Osborne %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-alvi19a %I PMLR %P 253--262 %U https://proceedings.mlr.press/v97/alvi19a.html %V 97 %X Batch Bayesian optimisation (BO) has been successfully applied to hyperparameter tuning using parallel computing, but it is wasteful of resources: workers that complete jobs ahead of others are left idle. We address this problem by developing an approach, Penalising Locally for Asynchronous Bayesian Optimisation on K Workers (PLAyBOOK), for asynchronous parallel BO. We demonstrate empirically the efficacy of PLAyBOOK and its variants on synthetic tasks and a real-world problem. We undertake a comparison between synchronous and asynchronous BO, and show that asynchronous BO often outperforms synchronous batch BO in both wall-clock time and sample efficiency.
APA
Alvi, A., Ru, B., Calliess, J., Roberts, S. & Osborne, M.A.. (2019). Asynchronous Batch Bayesian Optimisation with Improved Local Penalisation. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:253-262 Available from https://proceedings.mlr.press/v97/alvi19a.html.

Related Material