Research map / Optimization / Optimizers
VectorAdam for Rotation Equivariant Geometry Optimization
Selena Ling, Nicholas Sharp, Alec Jacobson
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 |
|---|---|---|---|
| replace per-coordinate second moments with scalar moments from each gradient vector's squared norm | Adam on vector-valued parameters · Laplacian regularization, 2D and 3D spherical meshes | rotation equivariance | Improved |
| use per-coordinate Adam updates on vector-valued parameters | Adam · Laplacian regularization, 2D and 3D spherical meshes | coordinate-system bias | Hurt |
| replace per-coordinate second moments with vector-norm second moments | Adam on 2D triangle mesh deformation · As-rigid-as-possible energy | coordinate-system bias | Improved |
| use per-coordinate Adam updates on vector-valued parameters | Adam on 2D triangle mesh deformation · As-rigid-as-possible energy | coordinate-system bias | Hurt |
| replace per-coordinate second moments with vector-norm second moments | Adam · Symmetric Dirichlet energy, rotated 2D mesh | rotation-consistent optimization path | Improved |
| replace per-coordinate second moments with vector-norm second moments | Adam · Symmetric Dirichlet energy, rotated 2D mesh | convergence rate | No clear effect |
| replace per-coordinate second moments with vector-norm second moments | Adam on point cloud inputs · Adversarial descent, ModelNet40 | gradient direction preservation | Improved |
| replace per-coordinate second moments with vector-norm second moments | Adam · Nefertiti inverse geometry reconstruction | Hausdorff distance | Improved |
| replace per-coordinate second moments with vector-norm second moments | Adam · Texture reconstruction | convergence rate | Improved |
| apply VectorAdam updates to rows of kernel gradients | PointNet · Point cloud segmentation, ShapeNet subset | training stability | Improved |
Closest papers by meaning
- Vector Neurons: A General Framework for SO(3)-Equivariant Networks (ICCV 2021 · similarity 0.88)
- Understanding Adam Requires Better Rotation Dependent Assumptions (NeurIPS 2025 · similarity 0.86)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.85)
- GradMetaNet: An Equivariant Architecture for Learning on Gradients (NeurIPS 2025 · similarity 0.84)
- Learning (Approximately) Equivariant Networks via Constrained Optimization (NeurIPS 2025 · similarity 0.84)
- Approximately Equivariant Neural Processes (NeurIPS 2024 · 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.
