Skip to content
arXiv · Statistics Machine Learning· Stephen Y Zhang, Gabriel Peyr\'e·· 4 hr agoAI score23

两层线性网络训练的全局指数收敛性

Global Exponential Convergence of Two-Layer Linear Network Training

AI brief

该研究证明,在 rich scaling 下以光滑 Polyak-Lojasiewicz 预测器损失训练的宽两层线性网络具有带显式速率的全局指数(线性)收敛性。

Source: arXiv · Statistics Machine Learning · arxiv.org