Kernel‐risk‐sensitive conjugate gradient algorithm with Student's‐t distribution based random fourier features

Kernel-risk-sensitive loss (KRSL) achieves an efficient performance surface, which has been applied in the kernel adaptive filters (KAFs) successfully. However, KRSL based KAFs use stochastic gradient descent (SGD) method optimization, usually suffer from inadequate accuracy with slow convergence speed. In this letter, conjugate is adopted optimization of function, and problem non-convexity addressed by twice half-quadratic (HQ) methods. For sparsification, a novel Student's-t distribution random Fourier feature (St-RFF) for improvement conventional RFF method. Thus, features kernel-risk-sensitive (St-RFFKRSCG) algorithm proposed. Simulations on Mackey-Glass time series prediction under non-Gaussian noises confirm superiorities terms performance, robustness, computational cost.