<p>In environments heavily influenced by outliers, the affine projection generalized maximum correntropy (APGMC) algorithm outperforms existing AP-like algorithms in robustness and effectiveness. Generally, APGMC achieves fast convergence with large step sizes while maintaining high filtering accuracy with smaller ones. To reconcile these strengths, we utilize a combined strategy incorporating two distinct approaches for the mixing parameter, i.e., the classical exponential method and the S-type function renowned for its computational efficiency. Simulation results validate the superiority of our proposed algorithms in system identification, particularly in environments challenged by heavily non-Gaussian impulsive noise.</p>