Closed-Form, Provable, and Robust PCA via Leverage Statistics and Innovation Search

The idea of Innovation Search, which was initially proposed for data clustering, recently used outlier detection. In the application Search detection, directions innovation were utilized to measure points. We study Values computed by algorithm under a quadratic cost function and it is proved that with new are equivalent Leverage Scores. This interesting connection establish several theoretical guarantees Score based robust PCA method design method. results include performance different models distribution outliers inliers. addition, we demonstrate robustness algorithms against presence noise. numerical studies indicate while presented approach fast closed-form, can outperform most existing algorithms.