学术报告（陈迪荣 9.25）

Abstract ：

In predicting multiple response variable from the predictor variable, the reduced-rank regression is an effective linear method by imposing that the matrix of regression coefficients is low rank. This talk considers a nonlinear version of reduced-rank regression with the help of kernel trick, which has been used extensively to extract nonlinear features of data. The optimal functions are obtained with the help of cross-covariance operators in reproducing kernel Hilbert spaces. Moreover, estimations of those optimal functions are constructed and the estimation errors are obtained. Simulation study is provided to illustrate the efficiency of the proposed method