Sotabase

Career

· Senior Staff Research Scientist, Google DeepMind2022–
· Associate Professor, Singapore University of Technology and Design (SUTD)2015–
· Wally Baer and Jeri Weiss Postdoctoral Scholar in Computing and Mathematical Sciences, Caltech
· PhD in Computer Science, Cornell University
· Postdoctoral Researcher, Georgia Institute of Technology
· Diploma in Electrical Engineering and Computer Science, National Technical University of Athens
· MSc in Applied Mathematics and Logic, National University of Athens
· Adjunct Faculty, Singapore University of Technology and Design (SUTD)
· Berkeley/Simons Fellow, UC Berkeley

Publications (189)

Cycles in adversarial regularized learning
ACM-SIAM Symposium on Discrete Algorithms · 2017
348
cited
Optimistic mirror descent in saddle-point problems: Going the extra (gradient) mile
International Conference on Learning Representations · 2018
321
cited
First-order methods almost always avoid strict saddle points
Mathematical programming · 2019
214
cited
Multiplicative updates outperform generic no-regret learning in congestion games: extended abstract
Symposium on the Theory of Computing · 2009
183
cited
The Unusual Effectiveness of Averaging in GAN Training
International Conference on Learning Representations · 2018
178
cited
Gradient Descent Only Converges to Minimizers: Non-Isolated Critical Points and Invariant Regions
Information Technology Convergence and Services · 2016
154
cited
Multiplicative Weights Update in Zero-Sum Games
ACM Conference on Economics and Computation · 2018
149
cited
Global Convergence of Multi-Agent Policy Gradient in Markov Potential Games
International Conference on Learning Representations · 2021
143
cited
Multiplicative Weights Update with Constant Step-Size in Congestion Games: Convergence, Limit Cycles and Chaos
Neural Information Processing Systems · 2017
140
cited
α-Rank: Multi-Agent Evaluation by Evolution
Scientific Reports · 2019
134
cited
First-order methods almost always avoid saddle points: The case of vanishing step-sizes
Neural Information Processing Systems · 2019
115
cited
From Poincaré Recurrence to Convergence in Imperfect Information Games: Finding Equilibrium via Regularization
International Conference on Machine Learning · 2020
99
cited
Incentives in Ethereum’s Hybrid Casper Protocol
International Conference on Blockchain · 2019
98
cited
Optimization Despite Chaos: Convex Relaxations to Complex Limit Sets via Poincaré Recurrence
ACM-SIAM Symposium on Discrete Algorithms · 2014
98
cited
No-regret learning and mixed Nash equilibria: They do not mix
Neural Information Processing Systems · 2020
93
cited
Multi-Agent Risks from Advanced AI
arXiv.org · 2025
88
cited
First-order Methods Almost Always Avoid Saddle Points
arXiv.org · 2017
84
cited
Beyond the Nash Equilibrium Barrier
International Conference on Supercomputing · 2011
80
cited
Dynamical analysis of the EIP-1559 Ethereum fee market
Conference on Advances in Financial Technologies · 2021
71
cited
Finite Regret and Cycles with Fixed Step-Size via Alternating Gradient Descent-Ascent
Annual Conference Computational Learning Theory · 2019
67
cited
Sotabase