Social and Information Sciences Laboratory (SISL) Seminar
Baxter 237
Two Proposals for Robust PCA using Semidefinite Programming
Joel Troop,
assistant professor of applied and computational mathematics,
Caltech,
The performance of principal component analysis (PCA) suffers badly in the presence of outliers. This talk discusses two novel approaches for robust PCA based on semidefinite programming. The first method, maximum mean absolute deviation rounding (MDR), seeks directions of large spread in the data while damping the effect of outliers. The second method produces a low-leverage decomposition (LLD) of the data by attempting to form a low-rank model for the data while separating out corrupted observations. Efficient computational methods for solving these SDPs are available, and numerical experiments confirm the value of these new techniques.
Joint work with Michael McCoy.
For more information, please contact Edith Quintanilla by phone at Ext. 3829 or by email at [email protected].