Iosif Sakos
Research Fellow at Singapore University of Technology and Design (SUTD)
My name is Iosif (aka Joseph). I am a Research Fellow at SUTD working with Antonios Varvitsiotis. My research lies at the intersection of algorithmic game theory and nonconvex optimization, with a primary focus on learning and equilibration in strategic multi-agent environments.
Research Overview
Multi-agent problems often admit a structural viewpoint that gives rise to nonconvex optimization formulations. For example:
- Extensive-form games are polynomial games in behavioral strategies.
- ML architectures such as GAN training can exhibit hidden-monotone structures.
My research develops optimization-based methods for certifying the existence and uniqueness of Nash equilibria, as well as the convergence rates of gradient-based algorithms in these settings. I further develop SDP-based frameworks for identifying and steering learning behavior in multi-agent systems, even when data are scarce or expensive to acquire. More recently, I have been working on exploiting an analogous structure in quantum ML, in particular on the development of provable approximation schemes for training PQCs.
Teaching Overview
I am actively involved in graduate-level teaching in optimization, games, and learning. In Spring 2025, I was a co-facilitator for Optimization for Data Science (with Antonios Varvitsiotis). In Fall 2025, I served as a teaching fellow for Special Topics in Games, Learning, and Optimization (with Anas Barakat and John Lazarsfeld), for which we developed a complete course website. My teaching emphasizes mathematical rigor, conceptual clarity, and the unifying role of optimization in learning and strategic interaction.