Research map / Probabilistic and scientific modeling / Dynamical systems
Unifying Predictions of Deterministic and Stochastic Physics in Mesh-reduced Space with Sequential Flow Generative Model
Luning Sun, Xu Han, Han Gao, Jian-Xun Wang, Liping Liu
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 |
|---|---|---|---|
| replace GMR-GMUS encoder with PbGMR-GMUS (residual connection + centers) (ablation) | GMR-GMUS · Cylinder flow | reconstruction error (x1e-3): 14.3 -> 1.9 | Improved |
| replace GMR-GMUS encoder with PbGMR-GMUS (ablation) | GMR-GMUS · Vascular flow | reconstruction error (x1e-3): 10 -> 2.8 | Improved |
| add centers to BFS graph autoencoder (no residual) (ablation) | graph autoencoder · Backward-facing step flow | reconstruction error v: 85 -> 19.2 | Improved |
| add residual connection to autoencoder with centers (ablation) | graph autoencoder · Backward-facing step flow | reconstruction error v: 19.2 -> 15.8 | Improved |
| add conditional normalizing flow to transformer-only variant with new encoder (ablation) | PbGMR-GMUS + transformer · Vascular flow | rollout error p: 3 -> 1.05 | Improved |
| add new encoder (residual+centers) to GMR-GMUS with CNF (ablation) | GMR-GMUS + CNF · Vascular flow | rollout error p: 9.2 -> 1.05 | Improved |
| replace 1-step conditional NF with attention-based temporal conditioning (ablation) | 1-step NF · Sonic flow | rollout error v: 16.36 -> 0.85 | Improved |
| replace LSTM sequence model with Transformer (ablation) | PbGMR-GMUS · Sonic flow | rollout error u: 10.3 -> 0.37 | Improved |
| reduce inference window length from 400 to 150 (ablation) | PbGMR-GMUS · Cylinder flow | rollout error p: 20 -> 22.99 | No clear effect |
| replace RealNVP with MAF flow (ablation) | PbGMR-GMUS · Sonic flow | rollout error u: 0.37 -> 0.33 | No clear effect |
| replace RealNVP with MAF flow (ablation) | PbGMR-GMUS · Vascular flow | rollout error u: 3.49 -> 3.82 | Hurt |
| increase number of centers up to 256 (ablation) | PbGMR-GMUS · Cylinder flow | recovery RMSE | Improved |
Closest papers by meaning
- Generating Physical Dynamics under Priors (ICLR 2025 · similarity 0.87)
- eXponential FAmily Dynamical Systems (XFADS): Large-scale nonlinear Gaussian state-space modeling (NeurIPS 2024 · similarity 0.87)
- Learning Stochastic Multiscale Models (NeurIPS 2025 · similarity 0.87)
- Space and time continuous physics simulation from partial observations (ICLR 2024 · similarity 0.87)
- Predicting partially observable dynamical systems via diffusion models with a multiscale inference scheme (NeurIPS 2025 · similarity 0.87)
- Probabilistic Forecasting with Stochastic Interpolants and Föllmer Processes (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.
