Reproduce “Uncertainty Quantification From Scaling Laws in Deep Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ibrahim Elsharkawy, Yonatan Kahn, Benjamin Hooberman
University of Illinois and University of Toronto researchers present a groundbreaking theoretical framework for quantifying uncertainty in neural networks through scaling laws, demonstrating that the coefficient of variation becomes independent of training set size in both infinite-width and finite-width networks while providing practical methods for efficient uncertainty estimation in scientific applications.
Key insights
- Both mean loss and variance follow scaling laws with training set size
- Coefficient of variation becomes independent of dataset size for sufficiently large datasets
- Infinite-width approximations can predict uncertainty behavior in practical finite-width networks
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.