Research map / Bandits and online learning / Multi-armed bandits
Thompson Sampling Guided Stochastic Searching on the Line for Deceptive Environments with Applications to Root-Finding Problems
Sondre Glimsdal, Ole-Christoffer Granmo
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 |
|---|---|---|---|
| increase door discretization |D| from 100 to 6400 (ablation) | TS-SPL · N-Door puzzle | convergence steps: 31.4 -> 40.9 | Hurt |
| increase truthfulness discretization |T| from 50 to 3200 (ablation) | TS-SPL · N-Door puzzle | convergence steps: 51.6 -> 52.1 | No clear effect |
| use correct prior over doors instead of flat (ablation) | TS-SPL · N-Door puzzle priors | convergence steps: 36.4 -> 30.2 | Improved |
| use correct prior over truthfulness instead of flat (ablation) | TS-SPL · N-Door puzzle priors | convergence steps: 36.4 -> 35.7 | Improved |
| use incorrect priors on both instead of flat (ablation) | TS-SPL · N-Door puzzle priors | convergence steps: 36.4 -> 113.1 | Hurt |
| learn truthfulness pi (TS-SPL) versus Bayesian-modified baselines in deceptive environment pi=0.15 | TS-SPL · Deceptive SPL | cumulative regret: 6.2 vs 352.5 | Improved |
| TS-SPL versus CPL-AdS in deceptive SPL | TS-SPL · Deceptive SPL (pi=0.85) | cumulative regret: 6.2 vs 501.6 | Improved |
| add Bayesian truthfulness model to NGBS (NGBS-M) in deceptive setting | NGBS · Deceptive SPL | cumulative regret: 1.4 -> 191.2 | Mixed |
| TS-SPL-INF (informative-only) versus TS-SPL at lambda*=0.95, pi=0.65 | TS-SPL · Informative SPL | average regret: 57.3 -> 23.9 | Improved |
| Bayesian schemes versus learning-automata schemes HSSL/SHSSL | TS-SPL · Informative SPL (lambda*=0.85, pi=0.65) | average regret: 36.4 vs 151.8 | Improved |
| TS-SPL versus SA on non-monotonic function B | TS-SPL · Root finding function B (pi=0.65) | average residual: 175.1 -> 47.1 | Improved |
| TS-SPL versus SA on monotonic function A | TS-SPL · Root finding function A (pi=0.65) | average residual: 32.4 vs 46.0 | Hurt |
Closest papers by meaning
- Minimal Exploration in Structured Stochastic Bandits (NeurIPS 2017 · similarity 0.90)
- Small steps no more: Global convergence of stochastic gradient bandits for arbitrary learning rates (NeurIPS 2024 · similarity 0.89)
- Thompson Sampling for Real-Valued Combinatorial Pure Exploration of Multi-Armed Bandit (AAAI 2024 · similarity 0.88)
- Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdown (NeurIPS 2025 · similarity 0.87)
- Feasible Action Search for Bandit Linear Programs via Thompson Sampling (ICML 2025 · similarity 0.87)
- Combinatorial Stochastic-Greedy Bandit (AAAI 2024 · similarity 0.87)
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.
