Fractional Gradient Descent-Based Auxiliary Model Algorithm for FIR Models with Missing Data

Abstract

This study proposes a fractional gradient descent (FGD) algorithm for FIR models with missing data. By using the auxiliary model method, the missing data can be obtained. Then, the FGD algorithm is applied to update the parameters of the FIR models. Because of the fractional term in the conventional GD algorithm, the convergence rates of the GD algorithm can be increased. In addition, to avoid the step-size calculation, an Aitken FGD-based auxiliary model algorithm is also introduced. The convergence analysis and simulation examples are provided to show the effectiveness of the proposed algorithms.