Bayesian Optimization with Resource Constraints and Production. Dolatnia, N., Fern, A., & Fern, X. In ![Bayesian Optimization with Resource Constraints and Production [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper abstract bibtex Bayesian optimization (BO) aims to optimize costly-to-evaluate functions by running a limited number of experiments that each evaluates the function at a selected input. Typical BO formulations assume that experiments are selected one at a time, or in fixed batches, and that experiments can be executed immediately upon request. This setup fails to capture many real-world domains where the execution of an experiment requires setup and preparation time, which may vary according to the type of experiment. In such domains, it is critical to explicitly plan for experiment preparation and setup activities in addition to making the usual BO decisions of which experiments to run. In this paper, we define a novel BO problem formulation that models the resources and activities needed to prepare and run experiments. We then present a planning approach, based on finite-horizon tree search, for scheduling the potentially current experimental activities with the aim of best optimizing the function within a limited time horizon. A key element of the approach is a novel state evaluation function for evaluating leaves of the search tree, for which we prove approximate guarantees. We evaluate the approach on a number of diverse benchmark problems and show that it produces high-quality results compared to a number of natural baselines.
@inproceedings {icaps16-204,
track = {Main Track},
title = {Bayesian Optimization with Resource Constraints and Production},
url = {http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13151},
author = {Nima Dolatnia and Alan Fern and Xiaoli Fern},
abstract = {Bayesian optimization (BO) aims to optimize costly-to-evaluate functions by running a limited number of experiments that each evaluates the function at a selected input. Typical BO formulations assume that experiments are selected one at a time, or in fixed batches, and that experiments can be executed immediately upon request. This setup fails to capture many real-world domains where the execution of an experiment requires setup and preparation time, which may vary according to the type of experiment. In such domains, it is critical to explicitly plan for experiment preparation and setup activities in addition to making the usual BO decisions of which experiments to run. In this paper, we define a novel BO problem formulation that models the resources and activities needed to prepare and run experiments. We then present a planning approach, based on finite-horizon tree search, for scheduling the potentially current experimental activities with the aim of best optimizing the function within a limited time horizon. A key element of the approach is a novel state evaluation function for evaluating leaves of the search tree, for which we prove approximate guarantees. We evaluate the approach on a number of diverse benchmark problems and show that it produces high-quality results compared to a number of natural baselines.},
keywords = {Scheduling under uncertainty,Probabilistic planning; MDPs and POMDPs}
}
Downloads: 0
{"_id":"sE4kdrDWBLGDXWZDr","bibbaseid":"dolatnia-fern-fern-bayesianoptimizationwithresourceconstraintsandproduction","downloads":0,"creationDate":"2016-03-09T03:04:32.878Z","title":"Bayesian Optimization with Resource Constraints and Production","author_short":["Dolatnia, N.","Fern, A.","Fern, X."],"year":null,"bibtype":"inproceedings","biburl":"icaps16.icaps-conference.org/papers.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","track":"Main Track","title":"Bayesian Optimization with Resource Constraints and Production","url":"http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13151","author":[{"firstnames":["Nima"],"propositions":[],"lastnames":["Dolatnia"],"suffixes":[]},{"firstnames":["Alan"],"propositions":[],"lastnames":["Fern"],"suffixes":[]},{"firstnames":["Xiaoli"],"propositions":[],"lastnames":["Fern"],"suffixes":[]}],"abstract":"Bayesian optimization (BO) aims to optimize costly-to-evaluate functions by running a limited number of experiments that each evaluates the function at a selected input. Typical BO formulations assume that experiments are selected one at a time, or in fixed batches, and that experiments can be executed immediately upon request. This setup fails to capture many real-world domains where the execution of an experiment requires setup and preparation time, which may vary according to the type of experiment. In such domains, it is critical to explicitly plan for experiment preparation and setup activities in addition to making the usual BO decisions of which experiments to run. In this paper, we define a novel BO problem formulation that models the resources and activities needed to prepare and run experiments. We then present a planning approach, based on finite-horizon tree search, for scheduling the potentially current experimental activities with the aim of best optimizing the function within a limited time horizon. A key element of the approach is a novel state evaluation function for evaluating leaves of the search tree, for which we prove approximate guarantees. We evaluate the approach on a number of diverse benchmark problems and show that it produces high-quality results compared to a number of natural baselines.","keywords":"Scheduling under uncertainty,Probabilistic planning; MDPs and POMDPs","bibtex":"@inproceedings {icaps16-204,\r\n track = {Main Track},\r\n title = {Bayesian Optimization with Resource Constraints and Production},\r\n url = {http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13151},\r\n author = {Nima Dolatnia and Alan Fern and Xiaoli Fern},\r\n abstract = {Bayesian optimization (BO) aims to optimize costly-to-evaluate functions by running a limited number of experiments that each evaluates the function at a selected input. Typical BO formulations assume that experiments are selected one at a time, or in fixed batches, and that experiments can be executed immediately upon request. This setup fails to capture many real-world domains where the execution of an experiment requires setup and preparation time, which may vary according to the type of experiment. In such domains, it is critical to explicitly plan for experiment preparation and setup activities in addition to making the usual BO decisions of which experiments to run. In this paper, we define a novel BO problem formulation that models the resources and activities needed to prepare and run experiments. We then present a planning approach, based on finite-horizon tree search, for scheduling the potentially current experimental activities with the aim of best optimizing the function within a limited time horizon. A key element of the approach is a novel state evaluation function for evaluating leaves of the search tree, for which we prove approximate guarantees. We evaluate the approach on a number of diverse benchmark problems and show that it produces high-quality results compared to a number of natural baselines.},\r\n keywords = {Scheduling under uncertainty,Probabilistic planning; MDPs and POMDPs}\r\n}\r\n\r\n","author_short":["Dolatnia, N.","Fern, A.","Fern, X."],"key":"icaps16-204","id":"icaps16-204","bibbaseid":"dolatnia-fern-fern-bayesianoptimizationwithresourceconstraintsandproduction","role":"author","urls":{"Paper":"http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13151"},"keyword":["Scheduling under uncertainty","Probabilistic planning; MDPs and POMDPs"],"metadata":{"authorlinks":{}},"downloads":0,"html":""},"search_terms":["bayesian","optimization","resource","constraints","production","dolatnia","fern","fern"],"keywords":["scheduling under uncertainty","probabilistic planning; mdps and pomdps"],"authorIDs":[],"dataSources":["iMkx859KiXcegwsin","EZtZjCTnxcdTTyeij"]}