?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Maximizing+acquisition+functions+for+Bayesian+optimization&rft.creator=Wilson%2C+JT&rft.creator=Hutter%2C+F&rft.creator=Deisenroth%2C+MP&rft.description=Bayesian+optimization+is+a+sample-efficient+approach+to+global+optimization+that%0D%0Arelies+on+theoretically+motivated+value+heuristics+(acquisition+functions)+to+guide%0D%0Aits+search+process.+Fully+maximizing+acquisition+functions+produces+the+Bayes%E2%80%99%0D%0Adecision+rule%2C+but+this+ideal+is+difficult+to+achieve+since+these+functions+are+frequently+non-trivial+to+optimize.+This+statement+is+especially+true+when+evaluating%0D%0Aqueries+in+parallel%2C+where+acquisition+functions+are+routinely+non-convex%2C+highdimensional%2C+and+intractable.+We+first+show+that+acquisition+functions+estimated%0D%0Avia+Monte+Carlo+integration+are+consistently+amenable+to+gradient-based+optimization.+Subsequently%2C+we+identify+a+common+family+of+acquisition+functions%2C+including+EI+and+UCB%2C+whose+properties+not+only+facilitate+but+justify+use+of+greedy%0D%0Aapproaches+for+their+maximization.&rft.publisher=Neural+Information+Processing+Systems+(NIPS)&rft.contributor=Bengio%2C+S&rft.contributor=Wallach%2C+H&rft.contributor=Larochelle%2C+H&rft.contributor=Grauman%2C+K&rft.contributor=CesaBianchi%2C+N&rft.contributor=Garnett%2C+R&rft.date=2018-12-08&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A+Bengio%2C+S+and+Wallach%2C+H+and+Larochelle%2C+H+and+Grauman%2C+K+and+CesaBianchi%2C+N+and+Garnett%2C+R%2C+(eds.)+Proceedings+of+the+32nd+Conference+on+Neural+Information+Processing+Systems+(NIPS+2018).++++Neural+Information+Processing+Systems+(NIPS)%3A+Montreal%2C+QC%2C+Canada.+(2018)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10083559%2F1%2F1805.10196v2.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10083559%2F&rft.rights=open