Research map / Probabilistic and scientific modeling / Variational and generative sampling
Stochastic Gradient Monomial Gamma Sampler
Yizhe Zhang, Changyou Chen, Zhe Gan, Ricardo Henao, Lawrence Carin
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 generalized monomial kinetics instead of Gaussian kinetics | SGMGT-D · Multiple-well synthetic potential | number of modes reached | Improved |
| Add Brownian/Langevin updates to theta and thermostat variables | SGMGT · Multiple-well synthetic potential | number of modes reached: 3 -> 5 | Improved |
| Use SGMGT-D instead of SGMGT (ablation) | Bayesian logistic regression · UCI datasets | AUROC: 0.92 -> 0.95; 0.78 -> 0.86; 0.91 -> 0.95; 0.86 -> 0.93; 0.87 -> 0.98; 0.70 -> 0.73 | Improved |
| Use monomial parameter a=2 rather than a=1 (ablation) | SGMGT-D · Bayesian logistic regression | AUROC: 0.95 -> 0.95; 0.86 -> 0.90; 0.95 -> 0.95; 0.93 -> 0.90; 0.98 -> 0.97; 0.73 -> 0.69 | Mixed |
| Use monomial parameter a=2 rather than a=1 (ablation) | SGMGT-D · Penn Treebank | perplexity: 113.8 -> 109.0 | Improved |
| Use monomial parameter a=2 rather than a=1 (ablation) | SGMGT · Penn Treebank | perplexity: 120.6 -> 250.5 | Hurt |
| Add stochastic resampling of momentum and thermostats (ablation) | SGMGT · MNIST DRBM | testing performance and mixing efficiency | Improved |
| Use SGMGT-D rather than SGMGT (ablation) | RNN sequence modeling · Polyphonic music and Penn Treebank | negative log-likelihood / perplexity: 7.90 -> 7.51; 4.35 -> 3.33; 8.42 -> 7.11; 8.67 -> 8.46; 120.6 -> 113.8 | Improved |
| Use SGMGT-D with a=2 rather than a=1 (ablation) | RNN sequence modeling · Polyphonic music and Penn Treebank | negative log-likelihood / perplexity: 7.51 -> 7.53; 3.33 -> 3.35; 7.11 -> 7.09; 8.46 -> 8.43; 113.8 -> 109.0 | Mixed |
| Use monomial parameter a=2 rather than a=1 (ablation) | SGMGT · ICML LDA test | perplexity: 986 -> 1013 | Hurt |
Closest papers by meaning
- Stochastic Gradient Richardson-Romberg Markov Chain Monte Carlo (NeurIPS 2016 · similarity 0.91)
- A Hybrid Stochastic Gradient Hamiltonian Monte Carlo Method (AAAI 2021 · similarity 0.91)
- Structured Stochastic Gradient MCMC (ICML 2022 · similarity 0.90)
- Parameter Expanded Stochastic Gradient Markov Chain Monte Carlo (ICLR 2025 · similarity 0.88)
- Faster Sampling via Stochastic Gradient Proximal Sampler (ICML 2024 · similarity 0.88)
- Parallelizing MCMC Across the Sequence Length (NeurIPS 2025 · similarity 0.87)
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.
