Research map / Optimization / Optimizers
Sampling in Constrained Domains with Orthogonal-Space Variational Gradient Descent
Ruqi Zhang, qiang liu, Xin T. Tong
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 |
|---|---|---|---|
| add orthogonal-space constraint to Langevin (SGLD -> O-Langevin) | SGLD · CIFAR10 ResNet-18 BNN | test error: 15.00 -> 4.46 | Improved |
| add orthogonal-space constraint to Langevin (SGLD -> O-Langevin) | SGLD · CIFAR10 ResNet-18 BNN | ECE: 2.21 -> 0.87 | Improved |
| add orthogonal-space constraint to Langevin (SGLD -> O-Langevin) | SGLD · CIFAR10 ResNet-18 BNN | AUROC on SVHN: 89.41 -> 98.68 | Improved |
| add orthogonal-space constraint to SVGD (SVGD -> O-SVGD) | SVGD · CIFAR10 ResNet-18 BNN | test error: 6.11 -> 4.92 | Improved |
| add orthogonal-space constraint to SVGD (SVGD -> O-SVGD) | SVGD · CIFAR10 ResNet-18 BNN | ECE: 0.93 -> 0.77 | Improved |
| add orthogonal-space constraint to SVGD (SVGD -> O-SVGD) | SVGD · CIFAR10 ResNet-18 BNN | AUROC on SVHN: 93.55 -> 94.69 | Improved |
| O-Langevin versus tempered SGLD | Tempered SGLD · CIFAR10 ResNet-18 BNN | test error: 4.73 -> 4.46 | Improved |
| O-Langevin versus tempered SGLD | Tempered SGLD · CIFAR10 ResNet-18 BNN | ECE: 0.83 -> 0.87 | Hurt |
| O-Langevin versus tempered SGLD | Tempered SGLD · CIFAR10 ResNet-18 BNN | AUROC on SVHN: 97.63 -> 98.68 | Improved |
| start sampler on manifold: O-SVGD and O-Langevin versus CLangevin and CHMC | CLangevin / CHMC · 2D synthetic constrained distribution | energy distance convergence | Improved |
| operate in ambient space without projection (MAE of |g(x)|) | CLangevin / CHMC · 2D synthetic constrained distribution | MAE |g(x)| | Hurt |
| O-Gradient versus standard Langevin and SVGD | Langevin / SVGD · Adult Income fairness | CV score (disparate impact) | Improved |
Closest papers by meaning
- Functional Gradient Flows for Constrained Sampling (NeurIPS 2024 · similarity 0.90)
- Sampling with Trusthworthy Constraints: A Variational Gradient Framework (NeurIPS 2021 · similarity 0.90)
- On the geometry of Stein variational gradient descent (JMLR 2023 · similarity 0.87)
- Guidance with Spherical Gaussian Constraint for Conditional Diffusion (ICML 2024 · similarity 0.85)
- No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths (ICML 2024 · similarity 0.85)
- Sobolev Regularized MMD Gradient Flow (NeurIPS 2026 · 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.
