Research map / Optimization / Implicit bias and generalization
SGD Converges to Global Minimum in Deep Learning via Star-convex Path
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 |
|---|---|---|---|
| Train MLP, Alexnet, and Inception with SGD to near-zero training loss | SGD · MNIST | iterationwise star-convexity | Helped |
| Train MLP, Alexnet, and Inception with SGD to near-zero training loss | SGD · CIFAR10 | iterationwise star-convexity | Helped |
| Train MLP, Alexnet, and Inception with SGD to near-zero training loss | SGD · MNIST | epochwise star-convexity | Helped |
| Train MLP, Alexnet, and Inception with SGD to near-zero training loss | SGD · CIFAR10 | epochwise star-convexity | Helped |
| Use an MLP with insufficient hidden neurons, resulting in training that does not reach zero loss (ablation) | SGD training on CIFAR10 · CIFAR10 | iterationwise star-convexity | Hurt |
| Use MSE loss instead of cross-entropy loss (ablation) | SGD training of neural networks · MNIST | epochwise star-convexity | Helped |
| Choose different intermediate near-zero-loss outputs as the reference point (ablation) | SGD training of Alexnet and MLP · CIFAR10 | epochwise star-convexity | No clear effect |
Closest papers by meaning
- Global Convergence and Stability of Stochastic Gradient Descent (NeurIPS 2022 · similarity 0.92)
- SGD Can Converge to Local Maxima (ICLR 2022 · similarity 0.91)
- Local SGD Converges Fast and Communicates Little (ICLR 2019 · similarity 0.91)
- Stochastic gradient descent performs variational inference, converges to limit cycles for deep networks (ICLR 2018 · similarity 0.91)
- An Alternative View: When Does SGD Escape Local Minima? (ICML 2018 · similarity 0.90)
- The Global Convergence Time of Stochastic Gradient Descent in Non-Convex Landscapes: Sharp Estimates via Large Deviations (ICML 2025 · similarity 0.90)
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.