Document Classification via Nonlinear Metric Learning
*延琴 白 (上海大学) email@example.com
In this talk, we propose a nonlinear metric learning method with multiple kernels. For solving the problem, we develop an intrinsic steepest descent algorithm with adaptive step-size on symmetric positive definite matrices groups. Finally, we show two experiments to demonstrate the performance the proposed nonlinear metric learning method and the steepest descent algorithm.