We consider unconstrained optimization problems where only “stochastic” estimates of the objective function are observable as replicates from a Monte Carlo simulation oracle. In the first study we assume that the function gradients are directly observable through the Monte Carlo simulation. We propose ASTRO, which is an adaptive sampling based trust-region optimization method where a stochastic local model is constructed, optimized, and updated iteratively. ASTRO is a derivative-based algorithm and provides almost sure convergence to a first-order critical point with good practical performance. In the second study the Monte Carlo simulation is assumed to provide no direct observations of the function gradient. We present ASTRO-DF, which is a class of derivative-free trust-region algorithms, where the stochastic local model is obtained through interpolation. Function estimation (as well as gradient estimation) and model construction within ASTRO and ASTRO-DF are adaptive in the sense that the extent of Monte Carlo sampling is determined by continuously monitoring and balancing metrics of sampling and structural errors within ASTRO and ASTRO-DF. Such error balancing is designed to ensure that the Monte Carlo effort within ASTRO and ASTRO-DF is sensitive to algorithm trajectory, sampling more whenever an iterate is inferred to be close to a critical point and less when far away. We demonstrate the almost-sure convergence of ASTRO-DF's iterates to a first-order critical point when using quadratic stochastic interpolation models. The question of using more complicated models, e.g., regression or stochastic kriging, in combination with adaptive sampling is worth further investigation and will benefit from the methods of proof we present. We investigate the implementation of ASTRO and ASTRO-DF along with the heuristics that enhance the implementation of ASTRO-DF, and report their finite-time performance on a series of low-to-moderate dimensional problems in the CUTEr framework. We speculate that the iterates of both ASTRO and ASTRO-DF achieve the canonical Monte Carlo convergence rate, although a proof remains elusive.
","abstract_html":"<p>We consider unconstrained optimization problems where only “stochastic” estimates of the objective function are observable as replicates from a Monte Carlo simulation oracle. In the first study we assume that the function gradients are directly observable through the Monte Carlo simulation. We propose ASTRO, which is an adaptive sampling based trust-region optimization method where a stochastic local model is constructed, optimized, and updated iteratively. ASTRO is a derivative-based algorithm and provides almost sure convergence to a first-order critical point with good practical performance. In the second study the Monte Carlo simulation is assumed to provide no direct observations of the function gradient. We present ASTRO-DF, which is a class of derivative-free trust-region algorithms, where the stochastic local model is obtained through interpolation. Function estimation (as well as gradient estimation) and model construction within ASTRO and ASTRO-DF are adaptive in the sense that the extent of Monte Carlo sampling is determined by continuously monitoring and balancing metrics of sampling and structural errors within ASTRO and ASTRO-DF. Such error balancing is designed to ensure that the Monte Carlo effort within ASTRO and ASTRO-DF is sensitive to algorithm trajectory, sampling more whenever an iterate is inferred to be close to a critical point and less when far away. We demonstrate the almost-sure convergence of ASTRO-DF's iterates to a first-order critical point when using quadratic stochastic interpolation models. The question of using more complicated models, e.g., regression or stochastic kriging, in combination with adaptive sampling is worth further investigation and will benefit from the methods of proof we present. We investigate the implementation of ASTRO and ASTRO-DF along with the heuristics that enhance the implementation of ASTRO-DF, and report their finite-time performance on a series of low-to-moderate dimensional problems in the CUTEr framework. We speculate that the iterates of both ASTRO and ASTRO-DF achieve the canonical Monte Carlo convergence rate, although a proof remains elusive.</p>","abstract_has_math":false,"creators":["Shashaani, Sara"],"institution":null,"degree_name":"Doctor of Philosophy (PhD)","degree_level":"Dissertation","degree_discipline":"Industrial Engineering","degree_department":null,"school":null,"contributors":["Raghu Pasupathy","Susan Hunter","Mohit Tawarmalani","Hong Wan"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2016,"date_issued":"2016-12-01T08:00:00Z","date_published":"2016-12-01T08:00:00Z","updated_at":"2026-07-24T03:54:09Z","subjects":["Applied sciences","Adaptive sampling","Derivative-free","Simulation","Trust-region","Operational Research"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://docs.lib.purdue.edu/open_access_dissertations/998","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Raghu Pasupathy","Susan Hunter","Mohit Tawarmalani","Hong Wan"]},{"key":"dc:creator","label":"Author","values":["Shashaani, Sara"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"thesis:degree_discipline","label":"Discipline","values":["Industrial Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Applied sciences","Adaptive sampling","Derivative-free","Simulation","Trust-region","Operational Research"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://docs.lib.purdue.edu/open_access_dissertations/998"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["We consider unconstrained optimization problems where only “stochastic” estimates of the objective function are observable as replicates from a Monte Carlo simulation oracle. In the first study we assume that the function gradients are directly observable through the Monte Carlo simulation. We propose ASTRO, which is an adaptive sampling based trust-region optimization method where a stochastic local model is constructed, optimized, and updated iteratively. ASTRO is a derivative-based algorithm and provides almost sure convergence to a first-order critical point with good practical performance. In the second study the Monte Carlo simulation is assumed to provide no direct observations of the function gradient. We present ASTRO-DF, which is a class of derivative-free trust-region algorithms, where the stochastic local model is obtained through interpolation. Function estimation (as well as gradient estimation) and model construction within ASTRO and ASTRO-DF are adaptive in the sense that the extent of Monte Carlo sampling is determined by continuously monitoring and balancing metrics of sampling and structural errors within ASTRO and ASTRO-DF. Such error balancing is designed to ensure that the Monte Carlo effort within ASTRO and ASTRO-DF is sensitive to algorithm trajectory, sampling more whenever an iterate is inferred to be close to a critical point and less when far away. We demonstrate the almost-sure convergence of ASTRO-DF's iterates to a first-order critical point when using quadratic stochastic interpolation models. The question of using more complicated models, e.g., regression or stochastic kriging, in combination with adaptive sampling is worth further investigation and will benefit from the methods of proof we present. We investigate the implementation of ASTRO and ASTRO-DF along with the heuristics that enhance the implementation of ASTRO-DF, and report their finite-time performance on a series of low-to-moderate dimensional problems in the CUTEr framework. We speculate that the iterates of both ASTRO and ASTRO-DF achieve the canonical Monte Carlo convergence rate, although a proof remains elusive.
"]},{"key":"dc:title","label":"Title","values":["Adaptive sampling trust-region methods for derivative-based and derivative-free simulation optimization problems"]}]}],"canonical_facts":{"dc:contributor":["Raghu Pasupathy","Susan Hunter","Mohit Tawarmalani","Hong Wan"],"dc:creator":["Shashaani, Sara"],"dc:description.abstract":["We consider unconstrained optimization problems where only “stochastic” estimates of the objective function are observable as replicates from a Monte Carlo simulation oracle. In the first study we assume that the function gradients are directly observable through the Monte Carlo simulation. We propose ASTRO, which is an adaptive sampling based trust-region optimization method where a stochastic local model is constructed, optimized, and updated iteratively. ASTRO is a derivative-based algorithm and provides almost sure convergence to a first-order critical point with good practical performance. In the second study the Monte Carlo simulation is assumed to provide no direct observations of the function gradient. We present ASTRO-DF, which is a class of derivative-free trust-region algorithms, where the stochastic local model is obtained through interpolation. Function estimation (as well as gradient estimation) and model construction within ASTRO and ASTRO-DF are adaptive in the sense that the extent of Monte Carlo sampling is determined by continuously monitoring and balancing metrics of sampling and structural errors within ASTRO and ASTRO-DF. Such error balancing is designed to ensure that the Monte Carlo effort within ASTRO and ASTRO-DF is sensitive to algorithm trajectory, sampling more whenever an iterate is inferred to be close to a critical point and less when far away. We demonstrate the almost-sure convergence of ASTRO-DF's iterates to a first-order critical point when using quadratic stochastic interpolation models. The question of using more complicated models, e.g., regression or stochastic kriging, in combination with adaptive sampling is worth further investigation and will benefit from the methods of proof we present. We investigate the implementation of ASTRO and ASTRO-DF along with the heuristics that enhance the implementation of ASTRO-DF, and report their finite-time performance on a series of low-to-moderate dimensional problems in the CUTEr framework. We speculate that the iterates of both ASTRO and ASTRO-DF achieve the canonical Monte Carlo convergence rate, although a proof remains elusive.
"],"dc:identifier":["https://docs.lib.purdue.edu/open_access_dissertations/998"],"dc:subject":["Applied sciences","Adaptive sampling","Derivative-free","Simulation","Trust-region","Operational Research"],"dc:title":["Adaptive sampling trust-region methods for derivative-based and derivative-free simulation optimization problems"],"thesis:degree_discipline":["Industrial Engineering"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-24T03:54:09Z"}