{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_52965"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_52965","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"NON-ADDITIVE MODELS FOR MULTI-FIDELITY SIMULATIONS","abstract":"Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2026","abstract_html":"Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2026","abstract_has_math":false,"creators":["Heo, Junoh"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Sung, Chih-Li C.-L","Hwang, Youngdeok Y","Sakhanenko, Lyudmila L","Weng, Haolei H"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T18:59:34Z","subjects":["Statistics"],"languages":["eng"],"rights":["Attribution 4.0 International"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52965","local:Heo_grad.msu_0128D_21998"],"render_values":[{"text":"etd:52965","href":null,"code":true},{"text":"local:Heo_grad.msu_0128D_21998","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/8hk5-qc43","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/8hk5-qc43","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_52965","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Sung, Chih-Li C.-L","Hwang, Youngdeok Y","Sakhanenko, Lyudmila L","Weng, Haolei H"]},{"key":"dc:creator","label":"Author","values":["Heo, Junoh"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Statistics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution 4.0 International"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52965","local:Heo_grad.msu_0128D_21998","https://doi.org/doi:10.25335/8hk5-qc43"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2026","Computer simulations have become indispensable tools for analyzing and optimizing complex systems across scientific and engineering disciplines. However, high-fidelity simulations often impose prohibitive computational costs, limiting their utility in iterative tasks. While multi-fidelity computer experiments have emerged as a powerful framework to balance accuracy and efficiency by integrating inexpensive low-fidelity data with costly high-fidelity data, conventional approaches predominantly rely on restrictive additive auto-regressive structures. This dissertation addresses these limitations by developing novel, flexible statistical models, scalable software implementations, and sequential design strategies for complex multi-fidelity emulation and optimization.First, this dissertation introduces the Recursive Non-Additive (RNA) emulator, a flexible Gaussian process-based model that recursively captures nonlinear and non-additive relationships across multi-fidelity data. By deriving closed-form expressions for the posterior predictive mean and variance, the RNA emulator achieves significant improvements in computational efficiency. Second, building upon the RNA framework, novel active learning strategies are proposed. These strategies dynamically optimize the selection of fidelity levels and input locations for subsequent simulation runs, effectively balancing emulation accuracy and computational costs.Third, the methodology is extended to address scenarios requiring precision beyond the highest available fidelity level. The Diffusion Non-Additive (DNA) model is proposed to extrapolate toward exact solutions when a tuning parameter controlling simulation accuracy is present but a zero-valued parameter remains computationally out of reach. Inspired by generative diffusion models, the DNA model employs a nonseparable covariance kernel to model complex interactions between the tuning parameter and input variables, improving both physical interpretability and predictive performance with closed-form uncertainty quantification.Fourth, to bridge the gap between theoretical advancement and practical application, \\textsf{RNAmf} and \\textsf{DNAmf} are introduced as open-source \\textsf{R} packages. These packages provide efficient, scalable tools for nested space-filling designs, multi-fidelity emulation, and active learning, making the proposed methodologies highly accessible to practitioners.Finally, this dissertation extends these advanced multi-fidelity emulators to Bayesian optimization. By leveraging the computationally efficient, non-additive framework and rigorous uncertainty quantification of the RNA model, novel cost-aware acquisition functions are developed. These sequential sampling strategies efficiently locate global optima in expensive black-box systems while actively minimizing total computational expenditure.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (170 p.) : ill."]},{"key":"dc:title","label":"Title","values":["NON-ADDITIVE MODELS FOR MULTI-FIDELITY SIMULATIONS"]}]}],"canonical_facts":{"dc:contributor":["Sung, Chih-Li C.-L","Hwang, Youngdeok Y","Sakhanenko, Lyudmila L","Weng, Haolei H"],"dc:creator":["Heo, Junoh"],"dc:date":["2026"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2026","Computer simulations have become indispensable tools for analyzing and optimizing complex systems across scientific and engineering disciplines. However, high-fidelity simulations often impose prohibitive computational costs, limiting their utility in iterative tasks. While multi-fidelity computer experiments have emerged as a powerful framework to balance accuracy and efficiency by integrating inexpensive low-fidelity data with costly high-fidelity data, conventional approaches predominantly rely on restrictive additive auto-regressive structures. This dissertation addresses these limitations by developing novel, flexible statistical models, scalable software implementations, and sequential design strategies for complex multi-fidelity emulation and optimization.First, this dissertation introduces the Recursive Non-Additive (RNA) emulator, a flexible Gaussian process-based model that recursively captures nonlinear and non-additive relationships across multi-fidelity data. By deriving closed-form expressions for the posterior predictive mean and variance, the RNA emulator achieves significant improvements in computational efficiency. Second, building upon the RNA framework, novel active learning strategies are proposed. These strategies dynamically optimize the selection of fidelity levels and input locations for subsequent simulation runs, effectively balancing emulation accuracy and computational costs.Third, the methodology is extended to address scenarios requiring precision beyond the highest available fidelity level. The Diffusion Non-Additive (DNA) model is proposed to extrapolate toward exact solutions when a tuning parameter controlling simulation accuracy is present but a zero-valued parameter remains computationally out of reach. Inspired by generative diffusion models, the DNA model employs a nonseparable covariance kernel to model complex interactions between the tuning parameter and input variables, improving both physical interpretability and predictive performance with closed-form uncertainty quantification.Fourth, to bridge the gap between theoretical advancement and practical application, \\textsf{RNAmf} and \\textsf{DNAmf} are introduced as open-source \\textsf{R} packages. These packages provide efficient, scalable tools for nested space-filling designs, multi-fidelity emulation, and active learning, making the proposed methodologies highly accessible to practitioners.Finally, this dissertation extends these advanced multi-fidelity emulators to Bayesian optimization. By leveraging the computationally efficient, non-additive framework and rigorous uncertainty quantification of the RNA model, novel cost-aware acquisition functions are developed. These sequential sampling strategies efficiently locate global optima in expensive black-box systems while actively minimizing total computational expenditure.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (170 p.) : ill."],"dc:identifier":["etd:52965","local:Heo_grad.msu_0128D_21998","https://doi.org/doi:10.25335/8hk5-qc43"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["Attribution 4.0 International"],"dc:subject":["Statistics"],"dc:title":["NON-ADDITIVE MODELS FOR MULTI-FIDELITY SIMULATIONS"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:34Z"}