BriefGPT.xyz
Sep, 2019
跨尺度泛化误差的建设性预测
A Constructive Prediction of the Generalization Error Across Scales
HTML
PDF
Jonathan S. Rosenfeld, Amir Rosenfeld, Yonatan Belinkov, Nir Shavit
TL;DR
本论文提出基于模型缩放的方法来构建适合各类模型和数据规模的函数形式，针对神经网络的泛化误差进行观测并给出了精确预测。
Abstract
The dependency of the
generalization error
of
neural networks
on model and dataset size is of critical importance both in practice and for understanding the theory of
→