Research map / Optimization / Implicit bias and generalization
The staircase property: How hierarchical structure can guide deep learning
Emmanuel Abbe, Enric Boix-Adserà, Matthew Stewart Brennan, Guy Bresler, Dheeraj Mysore Nagaraj
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| target the staircase function S_k instead of the single monomial chi_{1:k} | ResNet with SGD · Boolean hypercube n=30, k=10 | training error / Fourier coefficient recovery | Improved |
| target the staircase function instead of the parity monomial under unbiased Rademacher data | ResNet with SGD · n=30, k=10, m=6x10^4 | training error / Fourier coefficient recovery | Improved |
| target the staircase function instead of the monomial under Gaussian data | ResNet with SGD · n=100, k=5, m=3x10^5 | training error / Fourier coefficient recovery | Improved |
| target the staircase function instead of the monomial under biased product distribution p=0.75 | ResNet with SGD · n=30, k=7, m=6x10^4 | training error / Fourier coefficient recovery | Improved |
| learn the double staircase function S_{7,7} | ResNet with SGD · n=30, k=l=7, m=10^5 | training error / Fourier coefficient recovery | Improved |
| use ReLU ResNet with SGD instead of the theoretical quadratic-activation layerwise training | ResNet with SGD · Boolean hypercube staircase functions | training error / Fourier coefficient recovery | Improved |
Closest papers by meaning
- Learning Hierarchical Polynomials of Multiple Nonlinear Features (ICLR 2025 · similarity 0.86)
- Is Deeper Better only when Shallow is Good? (NeurIPS 2019 · similarity 0.86)
- Position: A Theory of Deep Learning Must Include Compositional Sparsity (ICML 2025 · similarity 0.85)
- The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents (ICML 2024 · similarity 0.85)
- No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths (ICML 2024 · similarity 0.85)
- Learning Hierarchical Polynomials with Three-Layer Neural Networks (ICLR 2024 · similarity 0.85)
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.
