Research map / Probabilistic and scientific modeling / Variational and generative sampling
Stochastic Variational Inference With Gradient Linearization
Tobias Plötz, Anne S. Wannenwetsch, Stefan Roth
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 gradient linearization for stochastic variational inference instead of gradient-based SVI with Adam | SVIGL · Sintel validation crops (100×100) | unnormalized KL: 4.24 -> 3.78 | Improved |
| Use gradient linearization for stochastic variational inference instead of gradient-based SVI with SGD | SVIGL · Sintel validation crops (100×100) | unnormalized KL: 4.45 -> 3.78 | Improved |
| Use gradient linearization for stochastic variational inference instead of Adam | SVIGL · BSDS test denoising | unnormalized KL: 0.98 -> 0.97 | No clear effect |
| Use gradient linearization for stochastic variational inference instead of Adam | SVIGL · BSDS test denoising | PSNR: 24.70 -> 24.77 | Improved |
| Use gradient linearization for stochastic variational inference instead of Adam | SVIGL · BSDS test denoising | SSIM: 0.680 -> 0.693 | Improved |
| Use SVIGL instead of MAP estimation with L-BFGS | Optical flow estimation · Sintel test | AEPE: 5.81 -> 5.74 | Improved |
| Vary the Adam step size (ablation) | SVI with Adam · Optical flow and Poisson-Gaussian denoising | unnormalized KL | Mixed |
| Reduce the sample set size (ablation) | SVI with Adam · Optical flow and Poisson-Gaussian denoising | unnormalized KL | Hurt |
| Reduce the sample set size (ablation) | SVIGL · Optical flow and Poisson-Gaussian denoising | convergence | No clear effect |
| Use Gaussian mean-field variational inference with SVIGL instead of a Laplace approximation | Poisson-Gaussian denoising · BSDS test | unnormalized KL: 1.57 -> 0.97 | Improved |
Closest papers by meaning
- Marginalized Stochastic Natural Gradients for Black-Box Variational Inference (ICML 2021 · similarity 0.91)
- Robust, Accurate Stochastic Optimization for Variational Inference (NeurIPS 2020 · similarity 0.91)
- Least squares variational inference (NeurIPS 2025 · similarity 0.89)
- Theoretical Guarantees for Variational Inference with Fixed-Variance Mixture of Gaussians (ICML 2024 · similarity 0.88)
- Natural Gradient VI: Guarantees for Non-Conjugate Models (NeurIPS 2025 · similarity 0.88)
- Variational Inference with Mixtures of Isotropic Gaussians (NeurIPS 2025 · similarity 0.88)
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.
