CAREER: Towards Data-Driven, Robust, and Versatile Online Resource Allocation

About This Grant

Online resource allocation lies at the core of computer science, operations research, and economics, addressing the persistent tradeoff between utilizing resources to seize immediate opportunities and conserving them for potentially better options in the future. Modern applications, such as online advertising, data center management, and ride-sharing operate on a massive scale, necessitating algorithms that learn from historical data, remain robust to outliers and corruptions, and adapt to diverse and dynamic scenarios. This project seeks to tackle these key technical challenges by developing innovative and effective solutions that will expand the boundaries of existing theory, address critical knowledge gaps, and pave the way for new innovations in online resource allocation. In addition to closely advising graduate students and developing curriculum, the educational components of this project include organizing a summer school on online algorithms, auction design, and data-driven algorithms, preparing a comprehensive textbook on Online Decision Making, and hosting a research workshop featuring invited talks from leading experts in academia and industry. The project focuses on three key directions in online resource allocation: (i) data-driven algorithms, which leverage limited historical data to infer problem structures and handle unknown input distributions; (ii) robust algorithms, designed to withstand adversarial corruptions, noise, and graphical correlations, ensuring resilience against errors in input models; and (iii) versatile algorithms, capable of addressing diverse objectives and constraints across a wide range of industrial applications. Tackling these directions involves addressing fundamental questions in economics, convex geometry, and optimization, including the role of prices in resource allocation, the structure of norms defined by general convex bodies, and the design of simple/fast algorithms for stochastic convex optimization. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.