Universal scaling laws in the gradient descent training of neural networks

Abstract

Current theoretical results on optimization trajectories of neural networks trained by gradient descent typically have the form of rigorous but potentially loose bounds on the loss values. In the present work we take a different approach and show that the learning trajectory can be characterized by an explicit asymptotic at large training times. Specifically, the leading term in the asymptotic expansion of the loss behaves as a power law $L(t)\sim t^{-\xi }$ with exponent $\xi$ expressed only through the data dimension, the smoothness of the activation function, and the class of function being approximated. Our results are based on spectral analysis of the integral operator representing the linearized evolution of a large network trained on the expected loss. Importantly, the techniques we employ do not require specific form of a data distribution, for example Gaussian, thus making our findings sufficiently universal.

References in corpus (7)

• The Loss Surfaces of Multilayer Networks
• Explaining Neural Scaling Laws
• The large learning rate phase of deep learning: the catapult mechanism
• On the linearity of large non-linear models: when and why the tangent kernel is constant
• Gradient Dynamics of Shallow Univariate ReLU Networks
• Optimal Rates for Averaged Stochastic Gradient Descent under Neural Tangent Kernel Regime
• Dynamically Stable Infinite-Width Limits of Neural Classifiers

Cited by in corpus (2)

• Learning curves for Gaussian process regression with power-law priors and targets
• Learning Curves for SGD on Structured Features