Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Flat Latent Manifolds with VAEs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Regularize the decoder metric tensor to a scaled identity and use latent mixup | VHP-VAE · MOT16 | IDF1: 60.5 -> 64.2 | Helped |
| Regularize the decoder metric tensor to a scaled identity and use latent mixup | VHP-VAE · MOT16 | IDs: 693 -> 598 | Helped |
| Regularize the decoder metric tensor to a scaled identity and use latent mixup | VHP-VAE · Human motion | observation distance ratio: 1.23 ± 0.20 -> 1.02 ± 0.06 | Helped |
| Regularize the decoder metric tensor to a scaled identity and use latent mixup | VHP-VAE · MNIST | observation distance ratio: 1.13 ±0.22 -> 1.01 ± 0.08 | Helped |
| Remove latent mixup from the metric regularization (ablation) | VHP-FMVAE · Human motion | metric regularity in regions without data | Hurt |
| Remove the scaled identity target c²I from the metric regularization (ablation) | VHP-FMVAE · Human motion | flat latent manifold | Hurt |
| Increase the Jacobian regularization strength from η=300 to η=3000 (ablation) | VHP-FMVAE-SORT · MOT16 | IDF1: 63.7 -> 64.2 | Helped |
| Increase the Jacobian regularization strength from η=300 to η=3000 (ablation) | VHP-FMVAE-SORT · MOT16 | MOTA: 59.1 -> 59.1 | No clear effect |
Closest papers by meaning
- Graph Geometry-Preserving Autoencoders (ICML 2024 · similarity 0.90)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.88)
- Task-Induced Riemannian Metrics for Vision Transformer Feature Spaces (NeurIPS 2026 · similarity 0.87)
- Euclidean Embedding of Data Using Local Distances (NeurIPS 2026 · similarity 0.87)
- Geometry-Aware Similarity Metrics for Neural Representations on Riemannian and Statistical Manifolds (NeurIPS 2026 · similarity 0.87)
- Follow the Energy, Find the Path: Riemannian Metrics from Energy-Based Models (NeurIPS 2025 · 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.