题目: Matrix Regression and Its Applications to Face Recognition
报告人: 杨健 教授 南京理工大学
摘要: We first outline the existing regression methods and their applications to face recognition. Then, we present the concept of matrix regression, which uses nuclear norm to characterize the representation error in a 2D matrix form. The ADMM algorithm and its fast version are developed for solving the matrix regression problem. We further analyze the optimality of using nuclear norm for error characterization. We also discuss some extensions of matrix regression model and their applications to face recognition.