Study establishes Bayes-optimal neural scaling laws for extensive-width networks
A revised arXiv paper analyzes the information-theoretic limits of learning a one-hidden-layer teacher network with hierarchical features from noisy queries, identifying sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-18 05:00 · arXiv stat.ML
Sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks