Research map / Bandits and online learning / Online convex optimization and games
Joint Online Learning and Decision-making via Dual Mirror Descent
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 |
|---|---|---|---|
| Use Thompson Sampling instead of least squares when revenue and contexts have no added randomness | Dual Mirror Descent parameter-learning step · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.0,0.0) | relative average revenue (%): 43.2 -> 98.1 | Helped |
| Use Thompson Sampling with added revenue noise | Dual Mirror Descent parameter-learning step · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.5,0.0) | relative average revenue (%): 98.1 -> 2.3 | Hurt |
| Use Thompson Sampling rather than least squares when revenue noise is (0.1,0.0) | Dual Mirror Descent parameter-learning step · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.1,0.0) | relative average revenue (%): 51.2 -> 13.2 | Hurt |
| Add randomized perturbation to ridge regression when contexts are deterministic and revenue noise is (0.0,0.0) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.0,0.0) | relative average revenue (%): 44.9 -> 59.3 | Helped |
| Add randomized perturbation to ridge regression when contexts are deterministic and revenue noise is (0.1,0.0) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.1,0.0) | relative average revenue (%): 52.9 -> 63.2 | Helped |
| Add randomized perturbation to ridge regression when contexts are deterministic and revenue noise is (0.5,0.0) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.5,0.0) | relative average revenue (%): 65.0 -> 67.7 | Helped |
| Add randomized perturbation to ridge regression when contexts have randomness (0.1) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.0,0.1) | relative average revenue (%): 95.6 -> 95.5 | No clear effect |
| Add randomized perturbation to ridge regression when revenue noise is (0.1,0.1) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.1,0.1) | relative average revenue (%): 94.5 -> 94.4 | No clear effect |
| Add randomized perturbation to ridge regression when revenue noise is (0.5,0.1) | Ridge regression parameter learner · Linear contextual bandits, T = 10000, (d × n) = (50,50), (0.5,0.1) | relative average revenue (%): 84.9 -> 85.2 | No clear effect |
| Use least squares with randomness added to contexts rather than deterministic contexts | Dual Mirror Descent parameter-learning step · Linear contextual bandits, T = 10000, (d × n) = (50,50), revenue noise (0.0) | relative average revenue (%): 43.2 -> 91.4 | Helped |
| Use ridge regression with randomness added to contexts rather than deterministic contexts | Dual Mirror Descent parameter-learning step · Linear contextual bandits, T = 10000, (d × n) = (50,50), revenue noise (0.0) | relative average revenue (%): 44.9 -> 95.6 | Helped |
Closest papers by meaning
- Online Learning under Budget and ROI Constraints via Weak Adaptivity (ICML 2024 · similarity 0.90)
- Revisiting Projection-Free Online Learning with Time-Varying Constraints (AAAI 2025 · similarity 0.88)
- Online Episodic Convex Reinforcement Learning (ICML 2025 · similarity 0.88)
- Online Composite Optimization Between Stochastic and Adversarial Environments (NeurIPS 2024 · similarity 0.87)
- Optimal Algorithms for Online Convex Optimization with Adversarial Constraints (NeurIPS 2024 · similarity 0.87)
- No-Regret Learning Under Adversarial Resource Constraints: A Spending Plan Is All You Need! (NeurIPS 2025 · 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.