Research map / Reinforcement learning / MDPs and planning
PAC Reinforcement Learning with Rich Observations
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 |
|---|---|---|---|
| Combine reactive Q-values, realizability, and deterministic transitions | LSVEE in the Contextual Decision Process model · Contextual Decision Process | sample complexity: At most \otil\left(\frac{M H^6K^2}{\epsilon^3}\log(N/\delta)\log(1/\delta)\right); independent of observation-space size | Helped |
| Remove the reactive-value-function assumption (ablation) | Learning reactive policies with realizability in a general POMDP · General POMDP | sample complexity: \Omega(K^H/\epsilon^2) | Hurt |
| Remove realizability of Q* (ablation) | Learning reactive policies in a general POMDP · General POMDP | sample complexity: \Omega(K^H/\epsilon^2) | Hurt |
| Use only reactive policies, without restricting Q* to be reactive (ablation) | Learning in a general POMDP · General POMDP | sample complexity: \Omega(K^H/\epsilon^2) | Hurt |
| Use realizability alone, without restricting predictors to current observation and action (ablation) | Learning in a general POMDP · General POMDP | sample complexity: \Omega(K^H/\epsilon^2) | Hurt |
| Use LSVEE for contextual bandits rather than a specialized contextual-bandit method | Contextual bandits (H=1, M=1) · Contextual bandits | sample complexity: \otil(\frac{K^2}{\epsilon^3}\log(N/\delta)\log(1/\delta)) vs. \otil(\frac{K}{\epsilon^2}\log(N/\delta)) | Hurt |
Closest papers by meaning
- On Oracle-Efficient PAC RL with Rich Observations (NeurIPS 2018 · similarity 0.91)
- Kinematic State Abstraction and Provably Efficient Rich-Observation Reinforcement Learning (ICML 2020 · similarity 0.89)
- Reinforcement Learning from Partial Observation: Linear Function Approximation with Provable Sample Efficiency (ICML 2022 · similarity 0.89)
- Rich-Observation Reinforcement Learning with Continuous Latent Dynamics (ICML 2024 · similarity 0.89)
- Sample-Efficient Learning of POMDPs with Multiple Observations In Hindsight (ICLR 2024 · similarity 0.88)
- Oracle-Efficient Reinforcement Learning for Max Value Ensembles (NeurIPS 2024 · similarity 0.88)
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.