Research map / Bandits and online learning / Online convex optimization and games
Fast Rates for Bandit Optimization with Upper-Confidence Frank-Wolfe
Quentin Berthet, Vianney Perchet
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use lower-confidence gradient estimates to choose the Frank-Wolfe vertex | Bandit optimization with convex loss · C-smooth convex functions on Δ_K | expected approximation error: 4 \sqrt{\frac{3K \log(T)}{T}} | Improved |
| Assume the loss is minimized at a vertex and gradient component gaps are positive | UCB Frank-Wolfe · Convex functions minimized at a simplex vertex | expected approximation error: \frac{48 \log(T)}{T} \sum_{i \neq \star}\frac{1}{\Delta^{(i)}(L)} | Improved |
| Assume strong convexity and an interior minimum at distance at least η from the boundary | UCB Frank-Wolfe · C-smooth strongly convex functions on Δ_K | expected approximation error: c_1 \frac{\log^2(T)}{T} + c_2\frac{\log(T)}{T} + c_3 \frac{1}{T} | Improved |
| Use only bandit feedback for linear losses rather than full gradient information | Any algorithm · Linear forms on Δ_K | minimax expected approximation error: c \sqrt{K/T} | Hurt |
| Use mixed feedback where an action reveals information about another gradient coefficient | Any algorithm · 1-strongly convex functions on Δ_3 | minimax expected approximation error: c/T^{2/3} | Hurt |
Closest papers by meaning
- Online Continuous Submodular Maximization: From Full-Information to Bandit Feedback (NeurIPS 2019 · similarity 0.90)
- Beyond Primal-Dual Methods in Bandits with Stochastic and Adversarial Constraints (NeurIPS 2024 · similarity 0.89)
- Optimal regret algorithm for Pseudo-1d Bandit Convex Optimization (ICML 2021 · similarity 0.89)
- Small steps no more: Global convergence of stochastic gradient bandits for arbitrary learning rates (NeurIPS 2024 · similarity 0.89)
- Refining the Confidence Level for Optimistic Bandit Strategies (JMLR 2018 · similarity 0.89)
- Online Stochastic Linear Optimization under One-bit Feedback (ICML 2016 · similarity 0.89)
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.
