When Both Layers Learn: Training Dynamics of Representing Linear Models via ReLU Networks figure
AlphaXiv 中文概览(可滚动查看)