Adaptive system training based on minimum error entropy
Yan Wang, Guo Weiguang, Hanwei Guo · 2004
Supervised adaptive system learning based on minimum error entropy method is studied in this article. To measure the information contained in error samples, Renyi's entropy is estimated with Parzen windowing. While MEE suffers from the high computational burden, so a segmentation method is brought forward to release it. MLP training base on MEE is derived, and MEE training for signal prediction is compared with MSE method. Simulation results verify the effectiveness of MEE method.