Research map / Probabilistic and scientific modeling / Variational and generative sampling
Density estimation on low-dimensional manifolds: an inflation-deflation approach
Christian Horvat, Jean-Pascal Pfister
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 inflation-deflation instead of M-flow for density evaluation | MNIST digit 1 density model · MNIST digit 1 (1000 digits) | evaluation time: 17.5 hours -> less than 10 seconds | Improved |
| use IID instead of M-flow for OOD detection | MNIST digit 1 likelihood · MNIST OOD | area of intersection: 0.07 for both methods | No clear effect |
| add normal-space noise (NID) in place of full Gaussian noise (IID) (ablation) | inflation-deflation · circle and surfaces, KS statistic | range of usable sigma^2 | Improved |
| use NID/IID trained in data space instead of FOM trained in latent space | density estimation on manifolds · surfaces, KS statistic | KS statistic | No clear effect |
| use IID instead of NID (ablation) | inflation-deflation · manifold with peaked latent density (Figure C) | optimal KS | Improved |
| increase embedding dimension D from 2 to 20 for the circle (ablation) | NID and IID · circle in R^D, KS statistic | optimal KS | No clear effect |
Closest papers by meaning
- Denoising Normalizing Flow (NeurIPS 2021 · similarity 0.92)
- Learning Distributions on Manifolds with Free-Form Flows (NeurIPS 2024 · similarity 0.87)
- Injective flows for star-like manifolds (ICLR 2025 · similarity 0.87)
- Diffusion Models Encode the Intrinsic Dimension of Data Manifolds (ICML 2024 · similarity 0.85)
- Learning Distances from Data with Normalizing Flows and Score Matching (ICML 2025 · similarity 0.85)
- Residual Deep Gaussian Processes on Manifolds (ICLR 2025 · similarity 0.85)
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.
