Research map / Bandits and online learning / Contextual bandits and Thompson sampling
On Approximate Thompson Sampling with Langevin Algorithms
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Warm-start Langevin sampler from previous round's sample (instead of re-initializing from prior) | ULA / SGLD approximate Thompson sampling · multi-armed bandit regret | number of Langevin iterations N needed | Helped |
| Scale posterior by gamma_a = 1/(8 d_a kappa_a^3) (posterior scaling for exploration) | exact Thompson sampling · multi-armed bandit regret | expected regret bound | Helped |
| Subsampled stochastic gradients (SGLD) with constant batch size and constant iterations instead of full-gradient ULA | approximate Thompson sampling · multi-armed bandit regret | per-round computational cost | Helped |
| Decrease approximation error of posterior sample over time (TV error of order 1/n) versus constant error | Thompson sampling with approximate posterior samples · Gaussian two-arm counterexample | cumulative regret | Mixed |
Closest papers by meaning
- Thompson Sampling with Less Exploration is Fast and Optimal (ICML 2023 · similarity 0.92)
- Langevin Thompson Sampling with Logarithmic Communication: Bandits and Reinforcement Learning (ICML 2023 · similarity 0.92)
- Thompson Sampling For Combinatorial Bandits: Polynomial Regret and Mismatched Sampling Paradox (NeurIPS 2024 · similarity 0.90)
- The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models (AAAI 2024 · similarity 0.90)
- Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdown (NeurIPS 2025 · similarity 0.89)
- Achieving adaptivity and optimality for multi-armed bandits using Exponential-Kullback Leibler Maillard Sampling (NeurIPS 2026 · 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.