GLASSES: Relieving The Myopia Of Bayesian Optimisation

[edit]

Javier Gonzalez, University of Sheffield
Michael Osborne, University of Oxford
Neil D. Lawrence, University of Sheffield

in Proceedings of the Nineteenth International Workshop on Artificial Intelligence and Statistics 51, pp 790-799

Related Material

Abstract

We present GLASSES: Global optimisation with Look-Ahead through Stochastic Simulation and Expected-loss Search. The majority of global optimisation approaches in use are myopic, in only considering the impact of the next function value; the non-myopic approaches that do exist are able to consider only a handful of future evaluations. Our novel algorithm, GLASSES, permits the consideration of dozens of evaluations into the future. This is done by approximating the ideal look-ahead loss function, which is expensive to evaluate, by a cheaper alternative in which the future steps of the algorithm are simulated beforehand. An Expectation Propagation algorithm is used to compute the expected value of the loss. We show that the far-horizon planning thus enabled leads to substantive performance gains in empirical tests.


@InProceedings{gonzalez-glasses16,
  title = 	 {{GLASSES}: Relieving The Myopia Of {B}ayesian Optimisation},
  author = 	 {Javier Gonzalez and Michael Osborne and Neil D. Lawrence},
  booktitle = 	 {Proceedings of the Nineteenth International Workshop on Artificial Intelligence and Statistics},
  pages = 	 {790},
  year = 	 {2016},
  editor = 	 {Arthur Gretton and Cristian Robert},
  volume = 	 {51},
  address = 	 {Cadiz, Spain},
  month = 	 {00},
  publisher = 	 {PMLR},
  edit = 	 {https://github.com/lawrennd//publications/edit/gh-pages/_posts/2016-05-02-gonzalez-glasses16.md},
  url =  	 {http://inverseprobability.com/publications/gonzalez-glasses16.html},
  abstract = 	 {We present GLASSES: Global optimisation with Look-Ahead through Stochastic Simulation and Expected-loss Search. The majority of global optimisation approaches in use are myopic, in only considering the impact of the next function value; the non-myopic approaches that do exist are able to consider only a handful of future evaluations. Our novel algorithm, GLASSES, permits the consideration of dozens of evaluations into the future. This is done by approximating the ideal look-ahead loss function, which is expensive to evaluate, by a cheaper alternative in which the future steps of the algorithm are simulated beforehand. An Expectation Propagation algorithm is used to compute the expected value of the loss. We show that the far-horizon planning thus enabled leads to substantive performance gains in empirical tests.},
  crossref =  {Gretton:aistats16},
  key = 	 {Gonzalez:glasses16},
  linkpdf = 	 {http://jmlr.org/proceedings/papers/v51/gonzalez16b.pdf},
  OPTgroup = 	 {}
 

}
%T GLASSES: Relieving The Myopia Of Bayesian Optimisation
%A Javier Gonzalez and Michael Osborne and Neil D. Lawrence
%B 
%C Proceedings of the Nineteenth International Workshop on Artificial Intelligence and Statistics
%D 
%E Arthur Gretton and Cristian Robert
%F gonzalez-glasses16
%I PMLR	
%P 790--799
%R 
%U http://inverseprobability.com/publications/gonzalez-glasses16.html
%V 51
%X We present GLASSES: Global optimisation with Look-Ahead through Stochastic Simulation and Expected-loss Search. The majority of global optimisation approaches in use are myopic, in only considering the impact of the next function value; the non-myopic approaches that do exist are able to consider only a handful of future evaluations. Our novel algorithm, GLASSES, permits the consideration of dozens of evaluations into the future. This is done by approximating the ideal look-ahead loss function, which is expensive to evaluate, by a cheaper alternative in which the future steps of the algorithm are simulated beforehand. An Expectation Propagation algorithm is used to compute the expected value of the loss. We show that the far-horizon planning thus enabled leads to substantive performance gains in empirical tests.
TY  - CPAPER
TI  - GLASSES: Relieving The Myopia Of Bayesian Optimisation
AU  - Javier Gonzalez
AU  - Michael Osborne
AU  - Neil D. Lawrence
BT  - Proceedings of the Nineteenth International Workshop on Artificial Intelligence and Statistics
PY  - 2016/05/02
DA  - 2016/05/02
ED  - Arthur Gretton
ED  - Cristian Robert	
ID  - gonzalez-glasses16
PB  - PMLR	
SP  - 790
EP  - 799
L1  - http://jmlr.org/proceedings/papers/v51/gonzalez16b.pdf
UR  - http://inverseprobability.com/publications/gonzalez-glasses16.html
AB  - We present GLASSES: Global optimisation with Look-Ahead through Stochastic Simulation and Expected-loss Search. The majority of global optimisation approaches in use are myopic, in only considering the impact of the next function value; the non-myopic approaches that do exist are able to consider only a handful of future evaluations. Our novel algorithm, GLASSES, permits the consideration of dozens of evaluations into the future. This is done by approximating the ideal look-ahead loss function, which is expensive to evaluate, by a cheaper alternative in which the future steps of the algorithm are simulated beforehand. An Expectation Propagation algorithm is used to compute the expected value of the loss. We show that the far-horizon planning thus enabled leads to substantive performance gains in empirical tests.
ER  -

Gonzalez, J., Osborne, M. & Lawrence, N.D.. (2016). GLASSES: Relieving The Myopia Of Bayesian Optimisation. Proceedings of the Nineteenth International Workshop on Artificial Intelligence and Statistics 51:790-799