Direction-Magnitude Decomposition for Low-Rank Matrix Optimization: Faster Convergence and Saddle-to-saddle Dynamics figure
AlphaXiv 中文概览(可滚动查看)