Dynamic Limited Variable Step-Size Algorithm Based on the MSD Variation Cost Function

The steady-state and convergence performances are important indicators to evaluate adaptive algorithms. step-size affects these two directly. Many relevant scholars have also proposed some variable algorithms for improving performance. However, there still problems in existing algorithms, such as the insufficient theoretical analysis, imbalanced performance unachievable parameter. These influence actual of greatly. Therefore, we intend further explore an inherent relationship between key this paper. variation mean square deviation (MSD) is adopted cost function. Based on analyses derivations, a novel algorithm with dynamic limited function (DLF) was proposed. At same time, sufficient analysis conducted weight stability. tested typical many different environments. Both experimental result all verified that equips superior