Research map / Reinforcement learning / Visual and representation RL
Value Prediction Network
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 value prediction and abstract-state planning instead of observation-prediction planning | Collect · Collect stochastic | average reward: 7.55 -> 8.11 | Helped |
| Use value prediction and abstract-state planning instead of observation-prediction planning | Collect · Collect deterministic | average reward: 9.30 -> 9.29 | No clear effect |
| Add multi-step planning beyond one-step value estimation (ablation) | VPN on Collect · Collect | average reward | Helped |
| Increase planning depth (ablation) | VPN on Collect · Collect | average reward | Helped |
| Plan deeper at evaluation than the training planning depth when trained with enough prediction steps (ablation) | VPN on Collect · Collect deterministic | average reward | Helped |
| Increase training planning depth while evaluating with one-step planning (ablation) | VPN on Collect · Collect deterministic | average reward | Hurt |
| Use VPN instead of DQN with short-lookahead planning | Atari games · Atari | average score | Mixed |
| Train the outcome module on additional random-policy data (ablation) | VPN on Collect · Collect | average reward | Helped |
| Predict only rewards, discounts, and abstract-state values rather than future observations | VPN on Collect · Collect deterministic | learning speed | Helped |
| Use VPN rather than OPN for generalization to unseen stochastic environments | Collect · Collect stochastic, unseen environments | average reward | Helped |
Closest papers by meaning
- Foresee then Evaluate: Decomposing Value Estimation with Latent Future Prediction (AAAI 2021 · similarity 0.87)
- The Value-Improvement Path: Towards Better Representations for Reinforcement Learning (AAAI 2021 · similarity 0.86)
- Use the Online Network If You Can: Towards Fast and Stable Reinforcement Learning (ICLR 2026 · similarity 0.86)
- In value-based deep reinforcement learning, a pruned network is a good network (ICML 2024 · similarity 0.86)
- Stop Regressing: Training Value Functions via Classification for Scalable Deep RL (ICML 2024 · similarity 0.86)
- Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy Churn (NeurIPS 2024 · similarity 0.85)
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.