Focus Period lund 2026
Postdoctoral Researcher
École Polytechnique Fédérale de Lausanne – EPFL (Switzerland)
Kimon Antonakopoulos completed his Bachelor’s degree in mathematics at the National and Kapodistrian University of Athens, Greece, followed by a Master’s in Optimization and Game Theory at Sorbonne University (formerly Paris VI – UPMC) in France. He then earned his Ph.D. at the University of Grenoble Alpes, where he worked with Prof. P. Mertikopoulos and Prof. E. V. Belmega on Adaptive Algorithms for Optimization Beyond Lipschitz Requirements.
Currently, Kimon Antonakopoulos is a postdoctoral researcher at the Laboratory for Information and Inference Systems (LIONS) at EPFL (École Polytechnique Fédérale de Lausanne), working with Prof. Volkan Cevher. His research explores the intersection of game theory, convex optimization, and variational inequalities, with a focus on developing efficient and adaptive algorithms for large-scale decision-making and learning problems.
Presenting: Geometry-Aware Adaptivity: From Convex Optimization to Games
Adaptivity is a central theme in modern optimization, governing how algorithms respond to oracle complexity, regularity properties, and problem geometry. It underlies both optimal complexity guarantees and practical efficiency across a broad class of convex optimization problems.
In this talk, I distinguish two complementary notions of adaptivity: black-box adaptivity, in which algorithms sequentially infer problem structure through oracle interactions, and a priori adaptivity, which exploits known regularity and geometric information to accelerate convergence.
I present a unified framework that combines these paradigms for general convex minimization, yielding algorithms that are simultaneously oracle-optimal and geometry-aware. I conclude by discussing extensions to multi-agent problems with combinatorial structure.