Research map / Reinforcement learning / Policy gradient and actor-critic
Linear Feature Encoding for Reinforcement Learning
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 linear encoder (learned by Algorithm 1/2) instead of random projection for features from raw images | linear value function approximation on raw images · Inverted Pendulum | balancing steps and probability of success | Helped |
| Increase number of encoder updates tau from 1 to 2 (ablation) | linear encoder (Algorithm 1) · Inverted Pendulum | balancing steps and probability of success | Mixed |
| Use linear encoder instead of random projection on noisy MNIST-based blackjack raw states | linear encoder · Blackjack (kappa=10) | relative value error and action error | Helped |
| Increase kappa (number of candidate MNIST images per digit) from 1 to 10 (ablation) | linear encoder · Blackjack | relative value error and action error | Hurt |
| Use linear encoder with limited effective rank (SVD truncation) instead of indicator features on true state | linear encoder · Blackjack (kappa=1, <=6000 episodes) | relative value error and action error | Helped |
| Use linear encoder with indicator-like features at large sample sizes | linear encoder · Blackjack (kappa=1, >6000 episodes) | relative value error and action error | Hurt |
Closest papers by meaning
- The Value-Improvement Path: Towards Better Representations for Reinforcement Learning (AAAI 2021 · similarity 0.88)
- Rethinking Model-based, Policy-based, and Value-based Reinforcement Learning via the Lens of Representation Complexity (NeurIPS 2024 · similarity 0.87)
- Stop Regressing: Training Value Functions via Classification for Scalable Deep RL (ICML 2024 · similarity 0.87)
- Zero-Shot Reinforcement Learning via Function Encoders (ICML 2024 · similarity 0.86)
- Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy Churn (NeurIPS 2024 · similarity 0.86)
- $\mathcal{B}$-Coder: Value-Based Deep Reinforcement Learning for Program Synthesis (ICLR 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.