Research map / Reinforcement learning / MDPs and planning
Loss Dynamics of Temporal Difference Reinforcement Learning
Blake Bordelon, Paul Masset, Henry Kuo, Cengiz Pehlevan
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 |
|---|---|---|---|
| increase batch size B | TD learning with linear features · synthetic power-law MDP | value error plateau | Improved |
| increase discount factor gamma | TD learning · synthetic power-law MDP | value error plateau | Hurt |
| increase learning rate eta | TD learning · synthetic power-law MDP | value error plateau | Hurt |
| anneal learning rate as eta_0 n^-chi | TD learning · synthetic power-law MDP | value error | Improved |
| use very large annealing exponent chi | TD learning · synthetic power-law MDP | value error | Hurt |
| scale-based reward shaping (w_phi parallel to w_TD) | TD learning · synthetic MDP | value error dynamics | No clear effect |
| rotation-based reward shaping toward high-variance feature directions | TD learning · synthetic MDP | convergence timescale | Improved |
| dense reward instead of sparse reward | TD learning with place cell code · 2D exploration MDP | value error convergence | Improved |
Closest papers by meaning
- Towards Parameter-Free Temporal Difference Learning (ICML 2026 · similarity 0.90)
- Almost Sure Convergence of Linear Temporal Difference Learning with Arbitrary Features (NeurIPS 2026 · similarity 0.87)
- An Improved Finite-time Analysis of Temporal Difference Learning with Deep Neural Networks (ICML 2024 · similarity 0.87)
- Discerning Temporal Difference Learning (AAAI 2024 · similarity 0.87)
- Reinforcement Learning Under Latent Dynamics: Toward Statistical and Algorithmic Modularity (NeurIPS 2024 · similarity 0.87)
- On Double Descent in Reinforcement Learning with LSTD and Random Features (ICLR 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.
