Research map / Optimization / Optimizers
Accelerating Convergence of Replica Exchange Stochastic Gradient MCMC via Variance Reduction
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 |
|---|---|---|---|
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR10, ResNet20 | prediction accuracy (%): 94.56$\pm$0.23 -> 94.84$\pm$0.11 | Helped |
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR10, ResNet32 | prediction accuracy (%): 95.44$\pm$0.16 -> 95.62$\pm$0.09 | Helped |
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR10, ResNet56 | prediction accuracy (%): 96.15$\pm$0.17 -> 96.32$\pm$0.15 | Helped |
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR100, ResNet20 | prediction accuracy (%): 73.94$\pm$0.34 -> 74.83$\pm$0.18 | Helped |
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR100, ResNet32 | prediction accuracy (%): 76.38$\pm$0.23 -> 77.40$\pm$0.27 | Helped |
| reduce the variance of energy estimates using control variates | reSGHMC · CIFAR100, ResNet56 | prediction accuracy (%): 79.86$\pm$0.26 -> 80.62$\pm$0.22 | Helped |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR10, ResNet20 | prediction accuracy (%): 94.84$\pm$0.11 -> 94.91$\pm$0.10 | Helped |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR10, ResNet32 | prediction accuracy (%): 95.62$\pm$0.09 -> 95.64$\pm$0.13 | No clear effect |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR10, ResNet56 | prediction accuracy (%): 96.32$\pm$0.15 -> 96.36$\pm$0.16 | No clear effect |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR100, ResNet20 | prediction accuracy (%): 74.83$\pm$0.18 -> 75.02$\pm$0.19 | Helped |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR100, ResNet32 | prediction accuracy (%): 77.40$\pm$0.27 -> 77.58$\pm$0.21 | No clear effect |
| add cyclic learning rates to variance-reduced replica exchange (ablation) | VR-reSGHMC · CIFAR100, ResNet56 | prediction accuracy (%): 80.62$\pm$0.22 -> 80.50$\pm$0.25 | No clear effect |
Closest papers by meaning
- ACCELERATING NONCONVEX LEARNING VIA REPLICA EXCHANGE LANGEVIN DIFFUSION (ICLR 2019 · similarity 0.92)
- Constrained Exploration via Reflected Replica Exchange Stochastic Gradient Langevin Dynamics (ICML 2024 · similarity 0.91)
- Non-convex Learning via Replica Exchange Stochastic Gradient MCMC (ICML 2020 · similarity 0.91)
- Improved Convergence Rate of Stochastic Gradient Langevin Dynamics with Variance Reduction and its Application to Optimization (NeurIPS 2022 · similarity 0.90)
- Variance Reduction in Stochastic Gradient Langevin Dynamics (NeurIPS 2016 · similarity 0.90)
- Generalization of noisy SGD in unbounded non-convex settings (ICML 2025 · 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.