Research map / Optimization / Implicit bias and generalization
Stability Based Generalization Bounds for Exponential Family Langevin 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 |
|---|---|---|---|
| Use gradient discrepancy instead of gradient norms in the stability bound | EFLD stability-based generalization bound · Generalization-bound comparison | generalization bound tightness | Helped |
| Use O(1/n) rather than O(1/√n) sample dependence | EFLD stability-based generalization bound · Generalization-bound comparison | sample-size dependence: O(1/n) instead of O(1/\sqrt{n}) | Helped |
| Use a data-dependent gradient-discrepancy step-size constraint instead of a global-Lipschitz constraint | EFLD stability analysis · Step-size bound | maximum allowable step size: \frac{\sigma_t}{L} -> \frac{\sigma_{t}}{\Delta_{t}(\bar{S})} | Helped |
| Compare the proposed bound with Rodríguez-Gálvez et al.'s bound | EFLD stability-based generalization bound · Bound comparison in Figure 1 | generalization bound tightness | Helped |
| Use expected stability instead of uniform stability | Stability-based generalization analysis · Generalization-bound analysis | bound tightness | Helped |
Closest papers by meaning
- Time-Independent Information-Theoretic Generalization Bounds for SGLD (NeurIPS 2023 · similarity 0.93)
- Time-independent Generalization Bounds for SGLD in Non-convex Settings (NeurIPS 2021 · similarity 0.93)
- On Generalization Error Bounds of Noisy Gradient Methods for Non-Convex Learning (ICLR 2020 · similarity 0.92)
- Generalization of noisy SGD in unbounded non-convex settings (ICML 2025 · similarity 0.92)
- Generalization Bounds for Gradient Methods via Discrete and Continuous Prior (NeurIPS 2022 · similarity 0.91)
- Temperature is All You Need for Generalization in Langevin Dynamics and other Markov Processes (NeurIPS 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.