Research map / Bandits and online learning / Online convex optimization and games
Sequential Counterfactual Risk Minimization
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 sequentially updated logging policies instead of repeatedly sampling from the initial logging policy | CRM · Pricing | test loss: -5.325 -> -5.575 | Helped |
| Use sequentially updated logging policies instead of repeatedly sampling from the initial logging policy | CRM · Potential | test loss: -.594 -> -.726 | Helped |
| Use sequentially updated logging policies instead of repeatedly sampling from the initial logging policy | CRM · Scene | test loss: .233 -> .219 | Helped |
| Use sequentially updated logging policies instead of repeatedly sampling from the initial logging policy | CRM · Yeast | test loss: .362 -> .294 | Helped |
| Use sequentially updated logging policies instead of repeatedly sampling from the initial logging policy | CRM · TMC | test loss: .158 -> .146 | Helped |
| Use the proposed lambda-selection heuristic rather than a posteriori fixed lambda selection | SCRM · Pricing | test loss: -5.353 ± .178 -> -5.575 ± .036 | Hurt |
| Use the proposed lambda-selection heuristic rather than a posteriori fixed lambda selection | SCRM · Potential | test loss: -.716 ± .020 -> -.726 ± .001 | Hurt |
| Use the proposed lambda-selection heuristic rather than a posteriori fixed lambda selection | SCRM · Yeast | test loss: .294 ± .026 -> .299 ± .039 | Hurt |
| Use the proposed lambda-selection heuristic rather than a posteriori fixed lambda selection | SCRM · TMC | test loss: .146 ± .012 -> .164 ± .021 | Hurt |
| Use the proposed lambda-selection heuristic rather than a posteriori fixed lambda selection | SCRM · Pricing | test loss: -5.353 ± .178 -> -5.575 ± .036 | Helped |
| Replace IPS, SNIPS, or clipped IPS with IPS-IX | Counterfactual policy evaluation · Sinusoidal-loss shifted-Gaussian evaluation | variance | Helped |
| Use sequential deployments rather than CRM | Gaussian quadratic-loss example · Pricing-style synthetic example | sample size to reach test-loss threshold: 100×2^10 -> 100×2^8; 100×2^16 -> 100×2^9; > 100×2^22 -> 100×2^11 | Helped |
Closest papers by meaning
- Bayesian Counterfactual Risk Minimization (ICML 2019 · similarity 0.92)
- Efficient Counterfactual Learning from Bandit Feedback (AAAI 2019 · similarity 0.89)
- Regret minimization in Linear Bandits with offline data via extended D-optimal exploration. (NeurIPS 2026 · similarity 0.85)
- Bayesian Regret Minimization in Offline Bandits (ICML 2024 · similarity 0.85)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.85)
- Factored-Reward Bandits with Intermediate Observations (ICML 2024 · similarity 0.84)
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.