Integrating Machine Learning and Optimization with Applications in Public Health and Sustainability
Citation
Wang, Kai. 2023. Integrating Machine Learning and Optimization with Applications in Public Health and Sustainability. Doctoral dissertation, Harvard University Graduate School of Arts and Sciences.Abstract
The field of artificial intelligence (AI) has garnered increasing attention in the realms of public health and conservation due to its potential to characterize complex dynamics and facilitate difficult decision-making. My research focuses on developing AI solutions, utilizing machine learning and optimization techniques, to provide actionable decisions for deployment and create positive social impact. This endeavor necessitates the integration of new algorithmic and learning paradigms, combining machine learning techniques to extract knowledge from data and optimization techniques to leverage domain knowledge and scale up to larger problem sizes. In this thesis, I present methodological and theoretical contributions in the integration of optimization into machine learning problems, including supervised learning, online learning, and multi-agent systems, with the aim of improving learning performance and scalability by harnessing the knowledge encoded in optimization tasks. Notably, this thesis introduces the first decision-focused learning to integrate sequential problems into the learning pipeline to provide feedback from decision-making processes and significantly reduce computation costs, thus enabling applications in large-scale public health problems. The proposed algorithm has been successfully applied in a field study and deployment in a maternal and child health program, marking the first successful implementation of decision-focused learning in the real world. Currently, the proposed algorithm is used by over 100,000 beneficiaries in India to enhance engagement with health information and translate algorithmic contributions into tangible social impact.Terms of Use
This article is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAACitable link to this page
https://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37375798
Collections
- FAS Theses and Dissertations [6566]
Contact administrator regarding this item (to report mistakes or request changes)