Seminar: Responsible and Data Driven Combinatorial Optimization

Abstract

Combinatorial Optimization (CO) problems are central to many real-world applications, often involving large-scale and sensitive datasets. The way these problems are solved has far-reaching implications, highlighting the need for "Responsible Combinatorial Optimization," which emphasizes scalable, ethical, and privacy-preserving solutions.

In this talk, I will begin with a generic framework for expressing CO problems and exploring their fundamental properties. I will then discuss submodular functions, an important class of CO problems that arise in areas such as machine learning, graph theory, and economics. I will present efficient algorithms for submodular maximization with provable guarantees that address scalability and privacy concerns.

Finally, I will discuss new advancements in leveraging machine learning tools to develop more data-driven and scalable approaches for solving CO problems. These advancements bridge the gap between traditional algorithmic techniques and emerging ML-driven methods.

_Biography

Akbar Rafiey is currently a Visiting Assistant Professor of Computer Science and Engineering at New York University’s Tandon School of Engineering. Before, he was a Postdoctoral Researcher at the Halıcıoğlu Data Science Institute (HDSI) and EnCORE, the Institute for Emerging CORE Methods in Data Science, at the University of California San Diego (UCSD). He holds a PhD in Computer Science and M.Sc. in Mathematics, both from Simon Fraser University (SFU). 

His research lies at the intersection of mathematics, computer science, and machine learning, with a focus on combinatorial optimization, responsible optimization, and leveraging machine learning for scalable combinatorial optimization techniques. His work has been published in leading venues such as STOC, ICML, ICALP, and AAAI and has been recognized with the NSERC Alexander Graham Bell Canada Graduate Scholarship.To learn more about Akbar’s work, please visit his webpage.