AI Paper Atlas

Research map / Bandits and online learning / Online convex optimization and games

Regret Bounds without Lipschitz Continuity: Online Learning with Relative-Lipschitz Losses

Yihan Zhou, Victor Sanches Portella, Mark Schmidt, Nicholas Harvey

NeurIPS 2020 · Bandits and online learning › Online convex optimization and games · approach: Constrained and unconstrained OCO algorithms · 7 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Extend FTRL analysis to losses relatively Lipschitz with respect to its regularizerFTRL · OCO with relative-Lipschitz lossesregret: Regret_T(z) ≤ 2 L √(K(T+1))Improved
Require losses to be relatively strongly convex as well as relatively Lipschitz with respect to the same reference functionFTL · OCO with relative-Lipschitz and relative-strongly-convex lossesregret: L^2/(2M) (log(T) + 1)Improved
Stabilize dual iterates with γ_t = η_{t+1}/η_tDS-OMD · OCO with relative-Lipschitz lossesregret: Regret_T(z) ≤ 2L√(K(T+1))Improved
Use γ_t = 1, skipping dual stabilization, when losses are relatively strongly convexOMD · OCO with relative-Lipschitz and relative-strongly-convex lossesregret: L^2/(2M)(log T+1)Improved
Add a known nonnegative convex extra regularizer Ψ, with Ψ(x_1)=0FTRL · Composite OCO with relative-Lipschitz lossescomposite regret: Regret_T^Ψ(z) ≤ 2 L √(K(T+1))Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy