Research map / Bandits and online learning / RL theory and MDPs
Improved No-Regret Algorithms for Stochastic Shortest Path with Linear MDP
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 the improved, model-agnostic analysis of finite-horizon approximation | Efficient finite-horizon algorithm for linear SSP · Linear SSP | regret: \tilo{\sqrt{d^3B_\star^3K/\cmin}} -> \tilo{\sqrt{d^3B_\star^2\ensuremath{T_\star} K} + d^3B_\star\ensuremath{T_\star}} | Helped |
| Use the efficient finite-horizon approximation and LSVI-UCB-style algorithm | Linear SSP · Linear SSP | regret: \tildeo{K^{5/6}} -> \tilo{\sqrt{d^3B_\star^2\ensuremath{T_\star} K} + d^3B_\star\ensuremath{T_\star}} | Helped |
| Use the restarted finite-horizon algorithm with the longer horizon and two-stage analysis | Efficient finite-horizon algorithm for linear SSP · Linear SSP | expected regret: \bigO{ \frac{d^3B_\star^4}{\cmin^2\text{\rm gap}_{\min}}\ln^5\frac{dB_\star K}{\cmin}} | Helped |
| Use VA-GOPO with variance-aware confidence sets and global optimization | Linear SSP · Linear SSP | regret: \tildeo{d^{3.5}B_\star\sqrt{K} + d^7B_\star^2} | Helped |
| Use the finite-horizon approximation with zero terminal costs and adaptive doubling | Linear SSP without known B_star or T_star · Linear SSP | regret: \tilo{\sqrt{d^3B_\star^3 K/\cmin} + d^3B_\star^2/\cmin} | Helped |
| Use the new finite-horizon approximation analysis with a tabular finite-horizon algorithm | Tabular SSP · Tabular SSP | regret: \tilo{B_\star\sqrt{SAK}+B_\star S^2A} | Helped |
| Use non-stationary policies through finite-horizon approximation instead of learning stationary policies only | Efficient linear SSP algorithm · Linear SSP | regret: \tildeo{\sqrt{d^3B_\star^2\ensuremath{T_\star} K} + d^3B_\star\ensuremath{T_\star}} | Helped |
Closest papers by meaning
- Minimax Regret for Stochastic Shortest Path (NeurIPS 2021 · similarity 0.93)
- Stochastic Shortest Path: Minimax, Parameter-Free and Towards Horizon-Free Regret (NeurIPS 2021 · similarity 0.92)
- Sharp Gap-Dependent Variance-Aware Regret Bounds for Tabular MDPs (NeurIPS 2025 · similarity 0.87)
- Achieving Tractable Minimax Optimal Regret in Average Reward MDPs (NeurIPS 2024 · similarity 0.86)
- Efficient Constraint Generation for Stochastic Shortest Path Problems (AAAI 2024 · similarity 0.86)
- Regret Lower Bounds for Decentralized Multi-Agent Stochastic Shortest Path Problems (NeurIPS 2025 · similarity 0.86)
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.