Research map / Probabilistic and scientific modeling / Dynamical systems
Leveraging Gauge Freedom for Learning Non-Gradient Population Dynamics of Stochastic Systems
Jules Berman, Tobias Blickhan, Benjamin Peherstorfer
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 non-gradient velocity field with weak continuity loss instead of gradient potential (DICE) | NGIF · Vlasov-Poisson two-stream | relative error in electric energy: 0.070 -> 0.043 (NGIF-Curl) | Improved |
| use non-gradient velocity field instead of gradient potential (DICE) | NGIF · Vlasov-Poisson bump-on-tail | relative error in electric energy: 0.283 -> 0.143 (NGIF-Kin) | Improved |
| use non-gradient velocity field instead of gradient potential (DICE) | NGIF · Vlasov-Poisson strong Landau (6D) | relative error in electric energy: 0.735 -> 0.542 (Kin), 0.864 (Curl) | Mixed |
| parameterize potential and use its gradient in NGIF loss instead of free vector field (ablation) | NGIF · tracer particles in turbulent field | TV distance | Hurt |
| use divergence regularizer instead of kinetic or curl (ablation) | NGIF · tracer particles in turbulent field | particle density fidelity | Improved |
| use NGIF variants instead of gradient baselines | NGIF · tracer particles in turbulent field | TV distance | Improved |
| use curl regularization with non-gradient field on rotating Gaussians | NGIF-Curl · Gigli example A | qualitative field agreement | Improved |
| use divergence regularizer instead of kinetic/curl on Vlasov two-stream (ablation) | NGIF-Div · Vlasov-Poisson two-stream | particle trajectories | Improved |
| use kinetic vs curl regularizer (ablation) | NGIF · tracer particles | TV distance | No clear effect |
| apply NGIF-Kin in latent space of autoencoder for 128x128 vorticity | NGIF-Kin · high-dim turbulence | enstrophy over time | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DICE — DiCE: The Infinitely Differentiable Monte Carlo Estimator (ICML 2018) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Parametric model reduction of mean-field and stochastic systems via higher-order action matching (NeurIPS 2024 · similarity 0.90)
- Curly Flow Matching for Learning Non-gradient Field Dynamics (NeurIPS 2025 · similarity 0.89)
- Wasserstein residuals: Learning Gradient Flows from Population Dynamics (NeurIPS 2026 · similarity 0.88)
- NETS: A Non-equilibrium Transport Sampler (ICML 2025 · similarity 0.87)
- Learning non-equilibrium diffusions with Schrödinger bridges: from exactly solvable to simulation-free (NeurIPS 2025 · similarity 0.87)
- Sampling in Unit Time with Kernel Fisher-Rao Flow (ICML 2024 · 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.
