Combined multiple random features least mean square algorithm for online applications

The multikernel least mean square (MKLMS) algorithm is a classical of adaptive filters due to its simplicity. However, the linear growth network structure main challenge MKLMS. To address this issue, novel multiple random features (MRFLMS) proposed by approximating Gaussian kernels with method. In addition, combined weight transfer strategy adopted in MRFLMS develop another (CMRFLMS) alleviate influence step-size on filtering performance and convergence rate. CMRFLMS fixed dimensional can provide comparable faster rate than Simulations prediction synthetic real non-linear system identification illustrate superiorities from aspects accuracy, rate, tracking performance.