Research map / Optimization / Optimizers
Information-Theoretic Generalization Bounds for SGLD via Data-Dependent Estimates
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 data-dependent priors to forecast minibatch gradients from a subset of the training data | SGLD generalization bounds · MNIST, Fashion-MNIST, CIFAR-10 | squared gradient incoherence relative to squared gradient norm: between 100 and 10,000 times smaller | Helped |
| Increase the held-out subset size m | Data-dependent SGLD generalization bound · Empirical bound comparisons | bound tightness | Helped |
| Use the data-dependent bound instead of the compared prior SGLD bound | SGLD · Monte Carlo bound estimates | generalization error bound | Helped |
| Use the data-dependent bound at small learning rates and inverse temperatures | SGLD · Monte Carlo bound estimates across learning-rate and inverse-temperature schedules | generalization error bound | Helped |
Closest papers by meaning
- Time-Independent Information-Theoretic Generalization Bounds for SGLD (NeurIPS 2023 · similarity 0.95)
- Optimizing Information-theoretical Generalization Bound via Anisotropic Noise of SGLD (NeurIPS 2021 · similarity 0.94)
- Time-independent Generalization Bounds for SGLD in Non-convex Settings (NeurIPS 2021 · similarity 0.92)
- Generalization of noisy SGD in unbounded non-convex settings (ICML 2025 · similarity 0.92)
- Exactly Tight Information-theoretic Generalization Bounds via Binary Jensen-Shannon Divergence (ICML 2025 · similarity 0.88)
- 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.