Research map / Reinforcement learning / Policy gradient and actor-critic
Taylor TD-learning
Michele Garibbo, Maxime Robeyns, Laurence Aitchison
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 TD updates instead of sample-based TD updates | TD-learning · Control tasks except Pendulum (batch updates) | update variance | Improved |
| use Taylor TD updates instead of sample-based TD updates | TD-learning · Pendulum (4-dimensional) | update variance | No clear effect |
| measure Taylor TD variance at no training, 5000 and 50000 steps (ablation) | Taylor TD · Half-Cheetah-v2 | update variance | Improved |
| replace sample-based expected updates with Taylor expansion (TaTD3 vs Sample-based Expected-TD3) (ablation) | TaTD3 · Half-Cheetah, Walker2d, Humanoid | average return | Improved |
| replace sample-based expected updates with Taylor expansion (ablation) | TaTD3 · Humanoid-v2 | average return | Improved |
| replace sample-based expected updates with Taylor expansion (ablation) | TaTD3 · Pendulum-v1 | average return | No clear effect |
| compare TaTD3 with MAGE | TaTD3 · Hopper-v2, Walker2d-v2, Ant-v2 | average return | Improved |
| compare TaTD3 with MBPO | TaTD3 · Humanoid-v2 | average return | Improved |
| compare TaTD3 with Dyna-TD3 and SAC | TaTD3 · six continuous control tasks | average return | Improved |
| add state/action perturbation expected-update estimation (Sample-based Expected-TD3 vs Dyna-TD3) (ablation) | TD3 with model · continuous control | average return | Hurt |
| increase data dimension in toy example (ablation) | Taylor expansion vs sample-based · toy function approximation | benefit of Taylor expansion | Improved |
Closest papers by meaning
- Adaptive Temporal Difference Learning With Linear Function Approximation (TPAMI 2021 · similarity 0.92)
- Discerning Temporal Difference Learning (AAAI 2024 · similarity 0.90)
- Almost Sure Convergence of Linear Temporal Difference Learning with Arbitrary Features (NeurIPS 2026 · similarity 0.89)
- Finite Sample Analysis of Linear Temporal Difference Learning with Arbitrary Features (NeurIPS 2025 · similarity 0.87)
- Revisiting a Design Choice in Gradient Temporal Difference Learning (ICLR 2025 · similarity 0.87)
- MAD-TD: Model-Augmented Data stabilizes High Update Ratio RL (ICLR 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.
