Research map / Optimization / Implicit bias and generalization
Early Stopping in Deep Networks: Double Descent and How to Eliminate it
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 |
|---|---|---|---|
| Scale feature-specific stepsizes so individual bias-variance tradeoffs reach their minima at the same iteration | Early-stopped least squares · Gaussian linear regression | risk | Helped |
| Use a smaller stepsize for the second layer | Two-layer ReLU network · Synthetic linear-model data | empirical risk | Helped |
| Use a smaller initialization scale for the first layer | Two-layer ReLU network · Synthetic linear-model data | empirical risk | Helped |
| Decrease the fully connected layer's stepsize | 5-layer CNN · CIFAR-10 with 20% random label noise | test error | Helped |
| Decrease the stepsizes of the latter half of the layers | ResNet-18 · CIFAR-10 with 20% random label noise | test error | Helped |
| Use different stepsizes for differently scaled features rather than a shared stepsize | Early-stopped least squares · Gaussian linear regression | risk | Helped |
Closest papers by meaning
- Double Trouble in Double Descent: Bias and Variance(s) in the Lazy Regime (ICML 2020 · similarity 0.91)
- Unraveling the Enigma of Double Descent: An In-depth Analysis through the Lens of Learned Feature Space (ICLR 2024 · similarity 0.90)
- On Double Descent in Reinforcement Learning with LSTD and Random Features (ICLR 2024 · similarity 0.88)
- Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse (ICLR 2025 · similarity 0.87)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.87)
- DSD²: Can We Dodge Sparse Double Descent and Compress the Neural Network Worry-Free? (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.