Barnabé Monnot

Research in algorithmic game theory, large systems and cryptoeconomics with a data-driven approach.

Publications + Talks

Posts

Reviews

Teaching

News

About + CV

Limits and Limitations of No-regret Learning in Games

Our paper investigates the issue of no-regret learning dynamics and their convergence. We show that convergence to a Nash Equilibrium that preserves the no-regret property is possible, but takes a long (exponential) time. In light of this result, we seek to extend typical Price of Anarchy bounds to the set of Coarse Correlated Equilibria (CCE) – which no-regret dynamics converge to – by comparing the cost of the best CCE to the optimum.

See publication on [Cambridge]

Back to publications