Research map / Optimization / Nonconvex and smooth optimization
Revisiting Optimal Convergence Rate for Smooth and Non-convex Stochastic Decentralized Optimization
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 MG-DSGD instead of DeTAG | decentralized training on homogeneous CIFAR-10 · CIFAR-10 homogeneous, ResNet-18, momentum | accuracy: 94.40 ± 0.13 -> 94.57 ± 0.05 | Helped |
| Use MG-DSGD instead of DeTAG | decentralized training on homogeneous CIFAR-10 · CIFAR-10 homogeneous, ResNet-20, momentum | accuracy: 91.77 ± 0.25 -> 91.77 ± 0.09 | No clear effect |
| Use MG-DSGD instead of DeTAG | decentralized training on homogeneous CIFAR-10 · CIFAR-10 homogeneous, ResNet-18, no momentum | accuracy: 93.17 ± 0.18 -> 93.75 ± 0.12 | Helped |
| Use MG-DSGD instead of DeTAG | decentralized training on homogeneous CIFAR-10 · CIFAR-10 homogeneous, ResNet-20, no momentum | accuracy: 88.97 ± 0.11 -> 89.18 ± 0.08 | Helped |
| Use MG-DSGD instead of DeTAG | decentralized training on heterogeneous CIFAR-10 · CIFAR-10, α=0.1, ResNet-20 | accuracy: 58.77 ± 1.30 -> 59.42 ± 3.36 | Mixed |
| Use MG-DSGD instead of DeTAG | decentralized training on heterogeneous CIFAR-10 · CIFAR-10, α=1, ResNet-20 | accuracy: 84.82 ± 0.19 -> 84.91 ± 0.74 | No clear effect |
| Use MG-DSGD instead of DeTAG | decentralized training on heterogeneous CIFAR-10 · CIFAR-10, α=10, ResNet-20 | accuracy: 90.27 ± 0.08 -> 90.43 ± 0.04 | Helped |
| Use MG-DSGD instead of DeTAG | decentralized training on ImageNet · ImageNet, ResNet-50 | accuracy: 76.57 ± 0.02 -> 76.64 ± 0.03 | Helped |
| Add momentum (ablation) | MG-DSGD on homogeneous CIFAR-10 · ResNet-18 and ResNet-20 | accuracy: 93.75 ± 0.12 -> 94.57 ± 0.05; 89.18 ± 0.08 -> 91.77 ± 0.09 | Mixed |
Closest papers by meaning
- Lower Bounds and Optimal Algorithms for Non-Smooth Convex Decentralized Optimization over Time-Varying Networks (NeurIPS 2024 · similarity 0.89)
- Decentralized Convex Finite-Sum Optimization with Better Dependence on Condition Numbers (ICML 2024 · similarity 0.89)
- Decentralized Sum-of-Nonconvex Optimization (AAAI 2024 · similarity 0.88)
- An Online Optimization Perspective on First-Order and Zero-Order Decentralized Nonsmooth Nonconvex Stochastic Optimization (ICML 2024 · similarity 0.88)
- Achieving Linear Speedup and Near-Optimal Complexity for Decentralized Optimization over Row-stochastic Networks (ICML 2025 · similarity 0.87)
- Jointly Improving the Sample and Communication Complexities in Decentralized Stochastic Minimax Optimization (AAAI 2024 · 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.