The Data-Reusing Maximum Correlation Entropy Algorithm Based with Logarithmic Functions

The data-reusing maximum correlation entropy algorithm (DRMCC) exhibits good convergence characteristics in adaptive filters when the input data are correlated. To further enhance the estimation performance of this algorithm for sparse systems, a penalty term in the form of a logarithmic function is incorporated, leading to the proposal of a logarithmic functionbased data-reusing maximum correlation entropy algorithm (DRMCCLZA). The penalty term imposes varying degrees of zero-attraction for distinct weight coefficients, enabling the algorithm to adjust the magnitude of zero-attraction in real time, contingent on the fluctuations of the weight coefficients throughout the update iteration process. The theoretical analysis of the convergence of the DRMCCLZA algorithm is presented herewith. The results of the simulation experiments demonstrate that the algorithm exhibits superior performance compared to other similar algorithms in system identification scenarios. Additionally, the algorithm demonstrates enhanced robustness in the context of sparse systems.