作者: Jian Sun , Jiang Wu , Tao Song , LiHong Hu , KaiLu Shan
DOI: 10.1021/JP502096Y
关键词:
摘要: The neural network correction approach that was previously proposed to achieve the chemical accuracy for first-principles methods is further developed by a combination of Kennard–Stone sampling and Bootstrapping methods. As result, calculated heat formation improved further, moreover, error bar each result can be determined. An enlarged database (Chen/13), which contains total 539 molecules made common elements H, C, N, O, F, S, Cl, constructed divided into training (449 molecules) testing (90 data sets with method. Upon correction, mean absolute deviation (MAD) B3LYP/6-311+G(3df,2p) reduced from 10.92 1.47 kcal mol–1 14.95 1.31 sets, respectively. Furthermore, method, broadly used statistical employed assess neu...