关键词：稀疏主成分分析；二维对称主成分分析；区域共坐标上升算法；共坐标上升算法；文本数据分析；半定松弛
摘 要：In this dissertation, we first discuss several formulations for Sparse PCA, as well as the algorithms for solving the formulations and a few greedy methods. We then develop a block coordinate ascent algorithm for solving DSPCA with better dependence on problem size. We show that our algorithm converges much faster than the existing first-order algorithm in practice and demonstrate that our code can handle huge real data sets. We also demonstrate that Sparse PCA does bring more interpretability and hence gives rise to new interesting findings in various real data sets.