Research map / Probabilistic and scientific modeling / Variational and generative sampling
Delving Into Discrete Normalizing Flows on SO(3) Manifold for Probabilistic Rotation Modeling
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 |
|---|---|---|---|
| Use Mobius coupling plus quaternion affine transformation | Rotation normalizing flow · SYMSOL I and SYMSOL II | log likelihood: 6.39 -> 10.38; 9.86 -> 12.94 | Helped |
| Use a pretrained Matrix Fisher distribution as the base distribution | Rotation normalizing flow · Pascal3D+ | Acc@30°: 0.827 -> 0.863 | Helped |
| Add quaternion affine transformation to Mobius coupling (ablation) | Rotation normalizing flow · SYMSOL I (avg.) | log likelihood: 9.41 -> 10.38 | Helped |
| Add quaternion affine transformation to Mobius coupling (ablation) | Rotation normalizing flow · SYMSOL I (avg.) | spread: 1.60 -> 0.67 | Helped |
| Add Mobius coupling to quaternion affine transformation (ablation) | Rotation normalizing flow · SYMSOL I (avg.) | log likelihood: 1.96 -> 10.38 | Helped |
| Replace quaternion affine transformation with rotation only (ablation) | Mobius coupling flow · SYMSOL I (avg.) | log likelihood: 10.38 -> 10.06 | Hurt |
| Use unconstrained matrix parameters instead of LU parameterization (ablation) | Quaternion affine transformation · SYMSOL I (avg.) | log likelihood: 8.57 -> 10.38 | Helped |
| Use unconstrained matrix parameters instead of LU parameterization (ablation) | Quaternion affine transformation · ModelNet10-SO3 | Acc@15: 0.727 -> 0.760 | Helped |
| Use quaternion affine transformation rather than rotation-matrix affine transformation (ablation) | Rotation normalizing flow · ModelNet10-SO3 | Acc@15: 0.734 -> 0.760 | Helped |
| Constrain Mobius parameters with the √2/2 trick to permit linear combinations (ablation) | Mobius coupling flow · Cube synthetic distribution | expressivity | Mixed |
Closest papers by meaning
- Normalizing Flows on the Product Space of SO(3) Manifolds for Probabilistic Human Pose Modeling (CVPR 2024 · similarity 0.90)
- Categorical Flow Matching on Statistical Manifolds (NeurIPS 2024 · similarity 0.85)
- Flow Matching on General Geometries (ICLR 2024 · similarity 0.85)
- Direct Product Flow Matching: Decoupling Radial and Angular Dynamics for Few-Shot Adaptation (NeurIPS 2026 · similarity 0.84)
- Rectifying Categorical Flows on Statistical Manifolds for One-Step Generation (NeurIPS 2026 · similarity 0.84)
- Normalizing Flows are Capable Generative Models (ICML 2025 · similarity 0.84)
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.