Research map / Reinforcement learning / Policy gradient and actor-critic
Propagating Uncertainty in Reinforcement Learning via Wasserstein Barycenters
Alberto Maria Metelli, Amarildo Likmeta, Marcello Restelli
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 |
|---|---|---|---|
| use Wasserstein-barycenter Q-posteriors (G-WQL, P-WQL) instead of classic Q-learning with Boltzmann exploration | WQL · Taxi, Six Arms, River Swim | online average return | Improved |
| use WQL instead of Bootstrapped Q-learning with double estimator | WQL · tabular tasks (Taxi, Chain, Six Arms) | online average return | Improved |
| use WQL instead of BQL on River Swim | WQL · River Swim | online average return | No clear effect |
| use WQL instead of Delayed Q-learning | WQL · all tabular tasks | online average return | Improved |
| use WQL instead of MBIE-EB | WQL · Chain and RiverSwim vs SixArms | online average return | Mixed |
| pair optimistic exploration (OX) with optimistic estimator (OE) rather than mismatched estimator (ablation) | G-WQL · Six Arms | online average return | Improved |
| pair posterior sampling exploration (PX) with posterior estimator (PE) rather than OE (ablation) | G-WQL · Six Arms | online average return | Improved |
| choice of exploration strategy and maximum estimator (ablation) | WQL variants · tabular tasks | online average return | Mixed |
| use Particle DQN instead of Double DQN and Bootstrapped DQN | PDQN · Asterix, Enduro | offline average return | Improved |
| use Particle DQN instead of DDQN and BDQN on Breakout | PDQN · Breakout | offline average return | No clear effect |
| narrow the particle initialization interval (ablation) | PDQN · Atari | learning speed / final performance | Mixed |
Closest papers by meaning
- Q-Distribution guided Q-learning for offline reinforcement learning: Uncertainty penalized Q-value via consistency model (NeurIPS 2024 · similarity 0.88)
- Deterministic Uncertainty Propagation for Improved Model-Based Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Bayesian Exploration Networks (ICML 2024 · similarity 0.87)
- Epistemic Bellman Operators (AAAI 2025 · similarity 0.87)
- Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities (NeurIPS 2026 · similarity 0.87)
- Selective Uncertainty Propagation in Offline RL (AAAI 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.
