Research map / Probabilistic and scientific modeling / Variational and generative sampling
Latent SDEs on Homogeneous Spaces
Sebastian Zeng, Florian Graf, Roland Kwitt
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 the sphere-based latent SDE instead of the compared interpolation models | Latent SDE on the sphere · PhysioNet (2012) interpolation, subsampled time points | MSE: 3.38 -> 3.14 (6 min, 50%); 3.33 -> 3.06 (6 min, 60%); 3.32 -> 3.01 (6 min, 70%); 3.33 -> 2.94 (6 min, 80%); 3.35 -> 3.00 | Improved |
| Use the sphere-based latent SDE instead of mTAND-Full | Latent SDE on the sphere · PhysioNet (2012) interpolation, 1 min, 50%-90% observed | MSE: 4.19 -> 3.14; 4.02 -> 3.03; 4.16 -> 3.01; 4.41 -> 2.99; 4.80 -> 2.96 | Improved |
| Use the sphere-based latent SDE instead of mTAND-Full | Latent SDE on the sphere · Human Activity test | accuracy: 91.1 -> 90.6 | No clear effect |
| Use the sphere-based latent SDE instead of the strongest listed baseline | Latent SDE on the sphere · Pendulum regression test | MSE: 3.20 -> 3.84 | Mixed |
| Use the sphere-based latent SDE instead of ODE$^2$VAE-KL | Latent SDE on the sphere · Rotating MNIST test | MSE: 18.8 -> 11.5 | Improved |
| Reduce the latent dimension from n=16 to n=3 (ablation) | Latent SDE on the sphere · Human Activity | accuracy: 90.6 -> 90.2 | No clear effect |
| Use more than one Chebyshev polynomial for drift evolution instead of K=1 (ablation) | Latent SDE on the sphere · Rotating MNIST interpolation and extrapolation | interpolation MSE | Mixed |
| Use a constant drift with K=1 rather than a higher-order time-varying drift (ablation) | Latent SDE on the sphere · Rotating MNIST | extrapolation quality | Improved |
Closest papers by meaning
- SING: SDE Inference via Natural Gradients (NeurIPS 2025 · similarity 0.90)
- Variational Inference for SDEs Driven by Fractional Noise (ICLR 2024 · similarity 0.90)
- Closing the Approximation Gap in Simulation-free Latent SDEs (NeurIPS 2026 · similarity 0.90)
- Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (NeurIPS 2026 · similarity 0.88)
- In-Context Learning of Stochastic Differential Equations with Foundation Inference Models (NeurIPS 2025 · similarity 0.87)
- Disentangling Continuous-Time Latent Dynamics: Identifiability of Latent SDEs via Diffusion Shifts (NeurIPS 2026 · 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.
