Convex combinations of kernel adaptive filters

Multikernel adaptive filtering has recently attracted significant research interest due to its enhanced flexibility and adaptation performance over single-kernel methods. In this paper, we focus on convex combinations of two single-kernel adaptive filters, characterized by different convergence speeds and steady-state performances, in order to get the best of both. We consider online estimation using single-kernel adaptive filters that may use different algorithms and kernels. Simulation results illustrate the efficiency of our approach.