Research map / Reinforcement learning / Policy gradient and actor-critic
Randomized Prior Functions for Deep Reinforcement Learning
Ian Osband, John Aslanides, Albin Cassirer
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 |
|---|---|---|---|
| add a fixed randomized prior function to each bootstrapped ensemble member | BootDQN · Chain environments | episodes to learn | Improved |
| replace additive prior functions with L2 regularization on initial weights | BootDQN · Chain environments | episodes to learn | Hurt |
| add an additive prior function (ablation) | Bootstrapped DQN · Chain environments | largest problem solved before the episode limit | Improved |
| vary the L2 regularization scale (ablation) | BSR · Chain environments | largest problem solved before the episode limit | No clear effect |
| increase ensemble size (ablation) | BSP · Chain environments | largest problem solved before the episode limit | Improved |
| use small prior scales (ablation) | BSP · Modified cartpole swing-up | cumulative reward | Hurt |
| increase prior scale (ablation) | BSP · Modified cartpole swing-up | cumulative reward | Mixed |
| add randomized prior networks | Bootstrapped DQN · Modified cartpole swing-up | cumulative reward | Improved |
| add randomized prior functions | Bootstrapped DQN · Montezuma's Revenge | game score | Improved |
| use dropout or NoisyNet for exploration | DQN · Chain environments | episodes to learn | Hurt |
| use state-visit-count UCB bonuses instead of randomized value functions | DQN · Chain environments | episodes to learn | Hurt |
| train independent ensemble members without prior functions | neural ensemble · Unobserved regions in 1D regression | predictive uncertainty | Hurt |
Closest papers by meaning
- Sample Efficient Deep Reinforcement Learning via Uncertainty Estimation (ICLR 2022 · similarity 0.89)
- Network Randomization: A Simple Technique for Generalization in Deep Reinforcement Learning (ICLR 2020 · similarity 0.89)
- Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models (NeurIPS 2018 · similarity 0.88)
- Deep Exploration via Randomized Value Functions (JMLR 2019 · similarity 0.87)
- Uncertainty-Based Smooth Policy Regularisation for Reinforcement Learning with Few Demonstrations (NeurIPS 2025 · similarity 0.86)
- Regularized Parameter Uncertainty for Improving Generalization in Reinforcement Learning (CVPR 2024 · 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.
