Back to Search Start Over

Asymptotics of feature learning in two-layer networks after one gradient-step

Authors :
Cui, Hugo
Pesce, Luca
Dandi, Yatin
Krzakala, Florent
Lu, Yue M.
Zdeborová, Lenka
Loureiro, Bruno
Source :
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:9662-9695, 2024
Publication Year :
2024

Abstract

In this manuscript, we investigate the problem of how two-layer neural networks learn features from data, and improve over the kernel regime, after being trained with a single gradient descent step. Leveraging the insight from (Ba et al., 2022), we model the trained network by a spiked Random Features (sRF) model. Further building on recent progress on Gaussian universality (Dandi et al., 2023), we provide an exact asymptotic description of the generalization error of the sRF in the high-dimensional limit where the number of samples, the width, and the input dimension grow at a proportional rate. The resulting characterization for sRFs also captures closely the learning curves of the original network model. This enables us to understand how adapting to the data is crucial for the network to efficiently learn non-linear functions in the direction of the gradient -- where at initialization it can only express linear functions in this regime.

Details

Database :
arXiv
Journal :
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:9662-9695, 2024
Publication Type :
Report
Accession number :
edsarx.2402.04980
Document Type :
Working Paper