Optimal bandwidth selection for re-substitution entropy estimation

摘要: Abstract A new fusion approach of selecting an optimal bandwidth for re-substitution entropy estimator (RE) is presented in this study. When approximating the continuous with density estimation, two types errors will be generated: estimation error (type-I error) and (type-II error). These are all strongly dependent on undetermined bandwidths. Firstly, experimental conclusion based 24 typical probability distributions demonstrated that there some inconsistency between bandwidths associated these errors. Secondly, different measures type-I type-II derived. trade-off a fundamental potential property our proposed method called RE I + II . Thus, conducted solved. Finally, comparisons carried out to verify performance strategy. The discretization deemed necessary preprocessing technology calculation traditionally. So, nine mostly used unsupervised methods introduced give comparison their computational performances And, five most popular estimators approximation also plugged into comparisons: splitting data (SDE), cross-validation (CVE), m-spacing (mSE), m n -spacing (mnSE), nearest neighbor distance (NNDE). simulation studies show can obtain better among involved methods. Meanwhile, behaviors revealed comparative results. empirical analysis demonstrates more insensitive generalizable way entropy. makes it possible handy derived from given dataset.