“…Deep learning can automatically learn effective classification features from a large number of complex unlabeled data and has strong data classification and prediction capabilities (Schmidhuber, ). Deep brief network (DBN) was first proposed by Hinton in 2006 (Hinton & Salakhutdinov, ) and was extended to many areas such as speech recognition (Sarikaya, Hinton, & Ramabhadran, ; Zhu, Chen, Zhao, Zhou, & Zhang, ), handwriting number recognition (Hinton, ), plant leaf classification (Liu & Kan, ), and power load prediction (Dedinec, Filiposka, Dedinec, & Kocarev, ; Wang et al, ). Nowadays, some experts have introduced deep learning into spectroscopy.…”