Research map / Learning theory and algorithms / Kernels and regression theory
Towards Understanding Generalization via Decomposing Excess Risk Dynamics
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 |
|---|---|---|---|
| Decompose excess risk into noise-related VER and signal-related BER, applying stability bounds to VER and uniform convergence to BER | Generalization-bound analysis · Overparameterized linear regression | excess-risk bound | Helped |
| Decompose excess risk into VER and BER and bound each component separately | Generalization-bound analysis · Diagonal matrix recovery | excess-risk bound | Helped |
| Increase neural-network depth | Risk decomposition framework · Synthetic sparse linear-function recovery | decomposition factor a | Hurt |
| Increase neural-network width | Risk decomposition framework · Synthetic sparse linear-function recovery | decomposition quality | Helped |
| Use different optimizers (SGD, Adam, or Rprop) | Risk decomposition framework · Synthetic sparse linear-function recovery | decomposition robustness | No clear effect |
| Use variance training with ResNet-18 | Risk decomposition framework · CIFAR-10 | VER | Mixed |
Closest papers by meaning
- On the generalization of learning algorithms that do not converge (NeurIPS 2022 · similarity 0.90)
- Stability and Generalization Analysis of Gradient Methods for Shallow Neural Networks (NeurIPS 2022 · similarity 0.90)
- Generalizability of Neural Networks Minimizing Empirical Risk Based on Expressive Power (ICLR 2025 · similarity 0.90)
- Stability and Generalization of Learning Algorithms that Converge to Global Optima (ICML 2018 · similarity 0.89)
- Fantastic Generalization Measures are Nowhere to be Found (ICLR 2024 · similarity 0.88)
- Sharper Guarantees for Learning Neural Network Classifiers with Gradient Methods (ICLR 2025 · similarity 0.88)
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.