AI RESEARCH
On the Dynamic Regret of Following the Regularized Leader: Optimism with History Pruning
arXiv CS.LG
•
ArXi:2505.22899v3 Announce Type: replace We revisit the Follow the Regularized Leader (FTRL) framework for Online Convex Optimization (OCO) over compact sets, focusing on achieving dynamic regret guarantees. Prior work has highlighted the framework's limitations in dynamic environments due to its tendency to produce "lazy" iterates.