Research map / Optimization / Implicit bias and generalization
Global Convergence Analysis of Local SGD for Two-layer Neural Network without Overparameterization
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 |
|---|---|---|---|
| increase alpha from 1/8 to 1/4 | Local SGD · synthetic two-layer network | success probability: 0.93 -> 1.00 | Helped |
| increase alpha from 1/64 to 1/8 | Local SGD · synthetic two-layer network | success probability: 0.61 -> 0.93 | Helped |
| use local SGD instead of minibatch SGD at alpha = 1/8 | Local SGD · synthetic two-layer network | success probability: 0.91 -> 0.93 | No clear effect |
| use local SGD instead of minibatch SGD at alpha = 1/64 | Local SGD · synthetic two-layer network | success probability: 0.60 -> 0.61 | No clear effect |
| use local SGD instead of minibatch SGD from bad initial points | Local SGD · synthetic two-layer network, bad initialization | convergence speed of signal correction | Helped |
| small alpha (1/32, 1/16) with a fixed bad initial point | Local SGD and minibatch SGD · synthetic two-layer network | converged minimum | Hurt |
| alpha = 1/4 with the same bad initial point | Local SGD and minibatch SGD · synthetic two-layer network | converged minimum | Helped |
Closest papers by meaning
- Convergence Analysis of Two-layer Neural Networks with ReLU Activation (NeurIPS 2017 · similarity 0.89)
- SGD Converges to Global Minimum in Deep Learning via Star-convex Path (ICLR 2019 · similarity 0.89)
- Local SGD Converges Fast and Communicates Little (ICLR 2019 · similarity 0.88)
- LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging (NeurIPS 2026 · similarity 0.87)
- AdaGK-SGD: Adaptive Global Knowledge Guided Distributed Stochastic Gradient Descent (AAAI 2025 · similarity 0.86)
- A Precise Characterization of SGD Stability Using Loss Surface Geometry (ICLR 2024 · similarity 0.86)
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.