Research map / Probabilistic and scientific modeling / Variational and generative sampling
Structured Stochastic Gradient MCMC
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 structured dropout with fully factorized groups | pSGLD · CIFAR-10, SVHN, and FMNIST | IAC: 831 -> 620; 839 -> 803; 774 -> 737 | Helped |
| Use structured dropout with fully factorized groups | pSGLD · CIFAR-10, SVHN, and FMNIST | ESS: 6.89 -> 7.22; 6.82 -> 7.00; 7.16 -> 7.57 | Helped |
| Use structured dropout with fully factorized groups | pSGLD · CIFAR-10, SVHN, and FMNIST | accuracy: 92.43 -> 92.51; 94.72 -> 94.82; 91.14 -> 92.47 | Helped |
| Use structured dropout with fully factorized groups | SGHMC · CIFAR-10, SVHN, and FMNIST | accuracy: 92.35 -> 92.54; 94.73 -> 94.65; 92.81 -> 92.71 | Mixed |
| Use structured dropout with fully factorized groups | SGHMC · CIFAR-10 | IAC: 832 -> 751 | Helped |
| Use structured dropout with fully factorized groups | SGHMC · SVHN and FMNIST | IAC: 858 -> 827; 920 -> 923 | Mixed |
| Use structured dropout with fully factorized groups | SGHMC · CIFAR-10, SVHN, and FMNIST | ESS: 6.84 -> 6.26; 6.59 -> 6.56; 5.72 -> 5.70 | Mixed |
| Increase the number of parameter groups (ablation) | S-pSGLD · MNIST | accuracy | Hurt |
| Increase the number of parameter groups (ablation) | S-pSGLD · MNIST | mixing time (IAC/ESS) | Helped |
| Increase the number of groups (ablation) | S_d-pSGLD · MNIST | accuracy | No clear effect |
| Increase the number of groups (ablation) | S_d-pSGLD · MNIST | mixing time (IAC/ESS) | Helped |
| Use random parameter partitioning rather than other tested partition schemes (ablation) | S-SGLD · Regression datasets: Wine, Housing, Obesity, Bike, Concrete, Airfoil | MSE | Mixed |
Closest papers by meaning
- A Divergence Bound for Hybrids of MCMC and Variational Inference and an Application to Langevin Dynamics and SGVI (ICML 2017 · similarity 0.92)
- Stochastic Gradient Richardson-Romberg Markov Chain Monte Carlo (NeurIPS 2016 · similarity 0.92)
- Parameter Expanded Stochastic Gradient Markov Chain Monte Carlo (ICLR 2025 · similarity 0.92)
- Stochastic Gradient MCMC Methods for Hidden Markov Models (ICML 2017 · similarity 0.92)
- Stochastic Quasi-Newton Langevin Monte Carlo (ICML 2016 · similarity 0.92)
- Learning to Explore for Stochastic Gradient MCMC (ICML 2024 · similarity 0.89)
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.