Kernel least mean square based on conjugate gradient

Siyuan Peng, Zongze Wu, Wentao Ma, Badong Chen

Abstract: Kernel least mean square (KLMS) algorithm has been successfully applied in fields of adaptive filtering and online learning due to their ability to solve sequentially nonlinear problems by implicitly mapping the input signal to a high-dimensional reproducing kernel Hilbert space (RKHS). In this paper, we propose a novel adaptive algorithm called KLMS based on conjugate gradient (KLMS-CG), which uses the orthogonal search directions, instead of using the traditional steepest descent approach, to improve the convergence speed. Further, the quantized KLMS based on conjugate gradient (QKLMS-CG) is proposed to curb the growth of network. Simulation results indicate that the new algorithm can converge faster than the original KLMS while maintaining excellent accuracy.

External IDs:dblp:conf/icassp/PengWMC17