鲁棒主成分分析与增强拉格朗日乘子法.ppt

Robust PCA;Principal Component Analysis (PCA);;;Why use Robust PCA?;;;;Main Problem;;Difference from PCA;Principal Component Analysis (PCA) [E. J. Candès, Journal of ACM 58(1): 1-37];Probabilistic PCA;维数约减：PCA与Probabilistic PCA;ill-posed problem (1);ill-posed problem (2);Conditions for exact recovery/decomposition;低秩矩阵恢复问题;RPCA算法;;;罚函数法 Penalty Method;罚函数法 Penalty Method;拉格郎日乘子法Langrange Multiplier;增强拉格郎日乘子法Augmented Langrange Multiplier;ALM算法;Alternating Direction Method of Multipliers ADMM (an variant of ALM);;;;;RPCA的ALM方法;;RPCA via the inexact ALM method;核范数与L1范数优化问题;L1范数优化;核范数优化;秩优化