Research map / Reinforcement learning / Policy gradient and actor-critic
Taylor Expansion of Discount Factors
Yunhao Tang, Mark Rowland, Remi Munos, Michal Valko
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 Taylor-expansion Q-function estimates with PPO (PPO(K), K=1, η=0.01) | PPO · MuJoCo and Bullet continuous-control tasks | undiscounted cumulative reward | Improved |
| Increase the PPO discount factor to 0.999 | PPO · MuJoCo and Bullet continuous-control tasks | undiscounted cumulative reward | Hurt |
| Use Taylor-expansion update weighting (K=100) | PPO · MuJoCo and Bullet continuous-control tasks | undiscounted cumulative reward | Improved |
| Augment states with the time step to account for the evaluation time limit | PPO · MuJoCo and Bullet continuous-control tasks | undiscounted cumulative reward | Mixed |
| Use recursive Q-function-difference bootstraps corresponding conceptually to K=∞, without a smaller-K trade-off | Open-source Q-function estimation baseline · continuous-control tasks | performance | No clear effect |
| Increase Taylor expansion order K in sample-based value estimation (ablation) | Taylor expansion value estimator · Tabular MDP numerical example | relative error | Mixed |
| Increase noise in the base value estimates (ablation) | Taylor expansion value estimator · Tabular MDP numerical example | optimal expansion order K* | Hurt |
| Vary the mixing coefficient η for the Taylor-expansion advantage estimator (ablation) | PPO(K) · MuJoCo and Bullet continuous-control tasks | overall performance | Mixed |
| Vary expansion order K for Taylor-expansion update weighting (ablation) | PPO with Taylor-expansion update weighting · MuJoCo and Bullet continuous-control tasks | performance | No clear effect |
Closest papers by meaning
- Is Value Learning Really the Main Bottleneck in Offline RL? (NeurIPS 2024 · similarity 0.88)
- An Analytical Study of Utility Functions in Multi-Objective Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Enhancing Value Function Estimation through First-Order State-Action Dynamics in Offline Reinforcement Learning (ICML 2024 · similarity 0.88)
- Exploration-Driven Policy Optimization in RLHF: Theoretical Insights on Efficient Data Utilization (ICML 2024 · similarity 0.88)
- Revisiting Value Iteration: Unified Analysis of Discounted and Average-Reward Cases (NeurIPS 2026 · similarity 0.88)
- Reward Model Learning vs. Direct Policy Optimization: A Comparative Analysis of Learning from Human Preferences (ICML 2024 · 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.
