Research map / Probabilistic and scientific modeling / Dynamical systems
Learning to Emulate Chaos: Adversarial Optimal Transport Regularization
Gabriel Melo, Leonardo Santiago, Peter Y. Lu
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 |
|---|---|---|---|
| add WGAN adversarial OT regularization | FNO emulator · L96 (multi-traj, noisy) | spectral distance: 0.321 -> 0.049 | Improved |
| add Sinkhorn adversarial OT regularization | FNO emulator · KS (single-traj, noisy) | spectral distance: 0.351 -> 0.219 | Improved |
| add WGAN adversarial OT regularization | FNO emulator · KS (single-traj, clean) | L1 histogram error: 0.241 -> 0.153 | Improved |
| add Sinkhorn (2D-Conv) OT regularization | UNet emulator · Kolmogorov Flow (single-traj, noisy) | L1 histogram error: 0.660 -> 0.360 | Improved |
| add OT regularization (learned) | emulator · one-step RMSE across systems | RMSE: 0.388 -> 0.401 | Hurt |
| replace fixed summaries with learned WGAN summaries | Fixed OT · L96 (multi-traj, noisy) | spectral distance: 0.141 -> 0.049 | Improved |
| replace fixed summaries with learned Sinkhorn summaries | Fixed OT · KS (single-traj, clean) | L1 histogram error: 0.310 -> 0.190 | Improved |
| replace fixed summaries with learned summaries | Fixed OT · L96 (multi-traj, clean) | L1 histogram error: 0.055 -> 0.083 | Hurt |
| use summary dimension d=3 instead of d=1 (ablation) | WGAN · L96 (multi-traj, noisy) | L1 histogram error: 0.234 -> 0.149 | Improved |
| use 1D-Conv summary map instead of MLP (ablation) | WGAN · KS (single-traj, noisy) | L1 histogram error: 0.339 -> 0.259 | Improved |
| use 1D-Conv summary map instead of MLP (ablation) | WGAN · KS (single-traj, noisy) | spectral distance: 0.148 -> 0.198 | Hurt |
| add WGAN OT regularization | FNO emulator · L96 single-traj | leading Lyapunov exponent: 1.615 -> 2.336 | Improved |
Closest papers by meaning
- Thermalizer: Stable autoregressive neural emulation of spatiotemporal chaos (ICML 2025 · similarity 0.89)
- Chaos Meets Attention: Transformers for Large-Scale Dynamical Prediction (ICML 2025 · similarity 0.87)
- Neural Emulator Superiority: When Machine Learning for PDEs Surpasses its Training Data (NeurIPS 2025 · similarity 0.86)
- First-Order Trajectory Matching: Fast Ensemble Predictions of Chaotic, Turbulent, Stochastic Systems (NeurIPS 2026 · similarity 0.85)
- Hierarchical Implicit Neural Emulators (NeurIPS 2025 · similarity 0.85)
- Mechanism-Aware Ensemble Conditioning for Data-Limited Emulation of Extreme Events (NeurIPS 2026 · similarity 0.85)
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.
