Research map / Image generation / Fast diffusion sampling
A Heat Diffusion Perspective on Geodesic Preserving Dimensionality Reduction
Guillaume Huguet, Alexander Tong, Edward De Brouwer, Yanlei Zhang, Guy Wolf, Ian Adelstein, Smita Krishnaswamy
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 HeatGeo instead of shortest-path geodesic estimate | HeatGeo · Swiss roll (noise 1.0) | Pearson correlation: 0.497 ± 0.144 -> 0.702 ± 0.086 | Improved |
| Use HeatGeo instead of PHATE for distance recovery | HeatGeo · Tree | Pearson correlation: 0.766 ± 0.023 -> 0.822 ± 0.008 | Improved |
| Use HeatGeo instead of PHATE for clustering | HeatGeo · PBMC | homogeneity: 0.798 ± 0.012 -> 0.734 ± 0.009 | Hurt |
| Use HeatGeo instead of UMAP for clustering | HeatGeo · PBMC | homogeneity: 0.177 ± 0.037 -> 0.734 ± 0.009 | Improved |
| Use HeatGeo instead of UMAP for clustering | HeatGeo · Tree | aMI: 0.681 ± 0.086 -> 0.786 ± 0.051 | Improved |
| Use HeatGeo instead of UMAP for clustering | HeatGeo · Swiss roll | homogeneity: 0.810 ± 0.036 -> 0.820 ± 0.008 | No clear effect |
| Use HeatGeo instead of UMAP for held-out time point interpolation | HeatGeo · IPSC | EMD: 0.866 ± 0.058 -> 0.365 ± 0.056 | Improved |
| Use HeatGeo instead of UMAP for held-out time point interpolation | HeatGeo · Cite | EMD: 0.791 ± 0.045 -> 0.890 ± 0.046 | Hurt |
| Use HeatGeo instead of t-SNE for held-out time point interpolation | HeatGeo · Multi | EMD: 1.208 ± 0.087 -> 0.958 ± 0.044 | Improved |
| Use HeatGeo instead of Diffusion Maps for clustering | HeatGeo · Tree | homogeneity: 0.341 ± 0.103 -> 0.784 ± 0.051 | Improved |
| Increase noise level from 0.1 to 1.0 for Heat-Geo distance recovery | Heat-Geo · Swiss roll | PearsonR: 0.992 ± 0.003 -> 0.702 ± 0.086 | Hurt |
| Increase noise level from 0.1 to 1.0 for PHATE distance recovery | PHATE · Swiss roll | PearsonR: 0.621 ± 0.006 -> 0.457 ± 0.01 | Hurt |
Closest papers by meaning
- Efficient Diffusion Models for Symmetric Manifolds (ICML 2025 · similarity 0.89)
- Generative Modeling on Manifolds Through Mixture of Riemannian Diffusion Processes (ICML 2024 · similarity 0.89)
- Manifold Diffusion Fields (ICLR 2024 · similarity 0.88)
- Toward a Unified Geometry Understanding : Riemannian Diffusion Framework for Graph Generation and Prediction (NeurIPS 2025 · similarity 0.88)
- Diffusion Models Encode the Intrinsic Dimension of Data Manifolds (ICML 2024 · similarity 0.88)
- What Matters for Diffusion-Friendly Latent Manifold? Prior-Aligned Autoencoders for Latent Diffusion (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.
