Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Riemannian Diffusion Models on General Manifolds via Physics-Informed Neural Networks
Gyeonghoon Ko, Juho Lee
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 PINN heat-kernel surrogate instead of Varadhan/spectral approximation | RSGM · Earthquakes | test log-likelihood: 0.10 -> 0.24 | Improved |
| use PINN heat-kernel surrogate instead of Varadhan/spectral approximation | RSGM · Volcano | test log-likelihood: 3.51 -> 3.56 | No clear effect |
| use PINN heat-kernel surrogate | RSGM · Floods | test log-likelihood: -0.53 -> -0.47 | Improved |
| use PINN heat-kernel surrogate | RSGM · Synthetic SO(3) | test log-likelihood: 0.18 -> 0.21 | Improved |
| use PINN-based diffusion instead of SPD-DDPM | SPD-DDPM · NYC taxi SPD(10) | mean Frobenius distance: 6.15 -> 5.21 | Improved |
| use PINN-based diffusion instead of EDM (VE) | EDM (VE) · QM9 | validity (%): 98.13 -> 99.60 | Improved |
| use PINN-based diffusion instead of EDM (VE) | EDM (VE) · QM9 | novelty (%): 63.30 -> 41.70 | Hurt |
| use PINN-based diffusion instead of EDM (VE) | EDM (VE) · QM9 | uniqueness (%): 99.12 -> 84.61 | Hurt |
| use PINN-based diffusion instead of DiffeoCFM | DiffeoCFM · BNCI 2014-002 | AUC: 0.81 -> 0.75 | Hurt |
| use PINN-based diffusion instead of LowTriDDPM | LowTriDDPM · BNCI 2014-002 | AUC: 0.65 -> 0.75 | Improved |
| use PINN-based diffusion instead of DiffeoCFM | DiffeoCFM · BNCI 2015-001 | CAS-F1: 0.67 -> 0.56 | Hurt |
Closest papers by meaning
- Generative Modeling on Manifolds Through Mixture of Riemannian Diffusion Processes (ICML 2024 · similarity 0.92)
- Efficient Diffusion Models for Symmetric Manifolds (ICML 2025 · similarity 0.91)
- Improving the Euclidean Diffusion Generation of Manifold Data by Mitigating Score Function Singularity (NeurIPS 2025 · similarity 0.90)
- Manifold Diffusion Fields (ICLR 2024 · similarity 0.89)
- Learning Distributions on Manifolds with Free-Form Flows (NeurIPS 2024 · similarity 0.88)
- Diffusion Models Encode the Intrinsic Dimension of Data Manifolds (ICML 2024 · 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.
