Research map / Probabilistic and scientific modeling / Differential equations and PDEs
LD-EnSF: Synergizing Latent Dynamics with Ensemble Score Filters for Fast Data Assimilation with Sparse Observations
Pengpeng Xiao, Phillip Si, Peng Chen
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 |
|---|---|---|---|
| evolve latent dynamics with LDNet surrogate instead of full dynamics | LD-EnSF · Kolmogorov Flow | T_d (s): 10,829 -> 0.049 | Improved |
| evolve latent dynamics with LDNet surrogate instead of full dynamics | LD-EnSF · Tsunami Modeling | T_d (s): 211.30 -> 0.050 | Improved |
| evolve latent dynamics with LDNet surrogate instead of full dynamics | LD-EnSF · Atmospheric Modeling | T_d (s): 35,603 -> 0.070 | Improved |
| assimilate in a 52-dim LDNet latent space instead of 512-dim VAE latent space | LD-EnSF · Atmospheric Modeling | T_f (s): 2.42 -> 1.42 | Improved |
| skip decoding to full state at every assimilation step | LD-EnSF · Atmospheric Modeling | T_r (s): 35.08 -> 0.017 | Improved |
| use improved LDNet (shifted initial latent state, two-stage training, ResNet and Fourier reconstruction net) rather than original LDNet (ablation) | LDNet · Tsunami Modeling | relative RMSE of full-state approximation | Improved |
| add Fourier encoding and improved architecture to LDNet (ablation) | LDNet · Kolmogorov Flow | relative RMSE of full-state approximation | Improved |
| replace VAE-dyn (VAE with LSTM) latent dynamics by LDNet | LD-EnSF · Kolmogorov, Tsunami, Atmospheric | relative RMSE of full-state approximation | Improved |
| use VAE-dyn surrogate as forward model in Latent-EnSF | Latent-EnSF · Kolmogorov, Tsunami, Atmospheric | assimilation relative RMSE | Hurt |
| assimilate in latent space instead of full space | EnSF · Kolmogorov, Tsunami, Atmospheric | assimilation relative RMSE | Improved |
| raise observation noise from none to 5%, 10%, 20% (ablation) | LD-EnSF · Kolmogorov, Tsunami, Atmospheric | relative RMSE of states and parameters | Hurt |
| use nonstationary Gaussian instead of Gaussian observation noise (ablation) | LD-EnSF · Tsunami Modeling | Assimilation error: 0.053 -> 0.086 | Hurt |
Closest papers by meaning
- Latent-EnSF: A Latent Ensemble Score Filter for High-Dimensional Data Assimilation with Sparse Observation Data (ICLR 2025 · similarity 0.93)
- FlowDAS: A Stochastic Interpolant-based Framework for Data Assimilation (NeurIPS 2025 · similarity 0.87)
- Sparsely-Supervised Data Assimilation via Physics-Informed Schrödinger Bridge (NeurIPS 2026 · similarity 0.87)
- Learning semilinear neural operators: A unified recursive framework for prediction and data assimilation. (ICLR 2024 · similarity 0.86)
- On conditional diffusion models for PDE simulations (NeurIPS 2024 · similarity 0.86)
- Tensor-Var: Efficient Four-Dimensional Variational Data Assimilation (ICML 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.
