Research map / Reinforcement learning / Policy gradient and actor-critic
Accelerating Q-learning through Efficient Value-Sharing across Actions
Prabhat Nagarajan, Brett Daley, Martha White, Marlos C. Machado
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 the mean-expansion layer (k=n) to DQN | DQN · Atari-57 | IQM human-normalized score: matches DQN's best performance at just under 20M timesteps, within 40% of training time | Improved |
| add the mean-expansion layer to IQN | IQN · Atari-57 | IQM human-normalized score: reaches IQN's highest score at 33.75M timesteps | Improved |
| add the mean-expansion layer to tabular Q-learning (IBQ) | Q-learning · 5x5 stochastic gridworld | completed episodes: over 20% more episodes within 1k timesteps | Improved |
| use mean-scaling k=n instead of k=1 (ablation) | IB-DQN · Atari-57 | IQM human-normalized score | Improved |
| increase k beyond a moderate value with fixed step size (ablation) | IB-DQN · Atari and LunarLander-v3 | AUC of return | Hurt |
| add Dueling DQN architecture to DQN | DQN · Atari-57 | IQM human-normalized score | No clear effect |
| add the mean-expansion layer to DQN with RMSprop and Huber loss | DQN · Atari (55 games) | IQM human-normalized score | Improved |
| add the mean-expansion layer to DQN | DQN · Atari-57 | value overestimation: reduced in all games | Improved |
| add the mean-expansion layer to DQN | DQN · Atari-57 | relative action gap | Improved |
| add the mean-expansion layer (k=n) to DQN | DQN · Atari Defender | mean score: 9260.5 -> 22271.0 | Improved |
Closest papers by meaning
- Stochastic Q-learning for Large Discrete Action Spaces (ICML 2024 · similarity 0.88)
- Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities (NeurIPS 2026 · similarity 0.88)
- Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Stop Regressing: Training Value Functions via Classification for Scalable Deep RL (ICML 2024 · similarity 0.87)
- Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network (ICML 2025 · similarity 0.86)
- Actor-Free Continuous Control via Structurally Maximizable Q-Functions (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.
