Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Scalable Inference in SDEs by Direct Matching of the Fokker–Planck–Kolmogorov Equation
Arno Solin, Ella Maija Tamir, Prakhar Verma
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 Euler-Maruyama sampling with moment matching | latent SDE on rotating MNIST · Rotating MNIST | MSE: 0.046 -> 0.051 | Hurt |
| replace Euler-Maruyama sampling with linearization | latent SDE on rotating MNIST · Rotating MNIST | MSE: 0.046 -> 0.052 | Hurt |
| replace Euler-Maruyama sampling with moment matching | latent SDE on rotating MNIST · Rotating MNIST | NLPD (t=64): 33.0 -> 52.7 | Hurt |
| replace Euler-Maruyama sampling with linearization | latent SDE on rotating MNIST · Rotating MNIST | NLPD (t=64): 33.0 -> 54.5 | Hurt |
| use assumed density (linearization) instead of 1 Euler-Maruyama path, GPU | latent SDE · MOCAP wall-clock GPU | time (s): 37.1 -> 2.1 | Improved |
| use assumed density (linearization) instead of 200 Euler-Maruyama paths, CPU | latent SDE · MOCAP wall-clock CPU | time (s): 94.2 -> 1.8 | Improved |
| use moment matching instead of linearization, GPU | assumed density SDE · MOCAP wall-clock GPU | time (s): 2.1 -> 6.0 | Hurt |
| use moment matching instead of linearization, CPU | assumed density SDE · MOCAP wall-clock CPU | time (s): 1.8 -> 1.8 | No clear effect |
| use assumed density approximation instead of Latent SDE sampling | latent SDE · MOCAP test | test MSE: 4.03 -> 7.55 | Hurt |
| use assumed density approximation instead of Latent ODE | latent SDE · MOCAP test | test MSE: 5.98 -> 7.55 | Hurt |
| use assumed density approximation instead of ODE2VAE-KL | latent SDE · MOCAP test | test MSE: 8.09 -> 7.55 | Improved |
| Euler-Maruyama paths needed to match moment matching accuracy, d=10 vs d=500 | Euler-Maruyama · Benes SDE | number of trajectories: 25 -> 550 | Mixed |
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.
- ODE2VAE — ODE2VAE: Deep generative second order ODEs with Bayesian neural networks (NeurIPS 2019) · better on 1 of 1 shared result
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
- SDE Matching: Scalable and Simulation-Free Training of Latent Stochastic Differential Equations (ICML 2025 · similarity 0.90)
- SING: SDE Inference via Natural Gradients (NeurIPS 2025 · similarity 0.90)
- Efficient Training of Neural Stochastic Differential Equations by Matching Finite Dimensional Distributions (ICLR 2025 · similarity 0.89)
- In-Context Learning of Stochastic Differential Equations with Foundation Inference Models (NeurIPS 2025 · similarity 0.89)
- Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions (NeurIPS 2025 · similarity 0.88)
- Closing the Approximation Gap in Simulation-free Latent SDEs (NeurIPS 2026 · similarity 0.88)
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.
