Research map / Reinforcement learning / Visual and representation RL
State Sequences Prediction via Fourier Transform for Representation Learning
Mingxuan Ye, Yufei Kuang, Jie Wang, Rui Yang, Wengang Zhou, Houqiang Li, Feng Wu
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 SPF representations instead of raw states | SAC and PPO · six MuJoCo tasks | return | Improved |
| Use SPF representations instead of OFENet representations | SAC and PPO · five of six MuJoCo tasks | return | Improved |
| Add target encoder, predictor, and projection networks (ablation) | SPF with SAC · HalfCheetah-v2 | return | Improved |
| Preserve low- and high-frequency components in the prediction loss (ablation) | SPF with SAC · HalfCheetah-v2 | convergence rate | Improved |
| Remove the projection layer (ablation) | SPF with SAC · HalfCheetah-v2 | return | No clear effect |
| Use an MLP-DenseNet encoder block instead of MLP (ablation) | SPF with SAC · HalfCheetah-v2 | return | Improved |
| Concatenate raw state with the MLP block (ablation) | SPF with SAC · HalfCheetah-v2 | return | Improved |
| Use infinite-step state sequences in the frequency domain as the prediction target | SAC representation learning · HalfCheetah-v2 | sample efficiency | Improved |
| Use state prediction rather than reward prediction | SAC representation learning · HalfCheetah-v2 | sample efficiency | Improved |
| Predict finite N-step state sequences rather than single-step future state | SAC representation learning · HalfCheetah-v2 | performance | Hurt |
Closest papers by meaning
- Bridging State and History Representations: Understanding Self-Predictive RL (ICLR 2024 · similarity 0.86)
- Towards Principled Representation Learning from Videos for Reinforcement Learning (ICLR 2024 · similarity 0.85)
- Diffusion Spectral Representation for Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- Fourier Controller Networks for Real-Time Decision-Making in Embodied Learning (ICML 2024 · similarity 0.85)
- Kernel Representation Learning for Time Sequence: Algorithm, Theory, and Applications (AAAI 2025 · similarity 0.84)
- Fourier Head: Helping Large Language Models Learn Complex Probability Distributions (ICLR 2025 · similarity 0.84)
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.
