Seven deep learning algorithms for predicting magnetic impurity spectral functions and the corresponding data sets

The codes and the corresponding data file related to the article "Prediction of impurity spectrum function by deep learning algorithm" published in China Physics B. Two zip files are given here, which are the data set of machine learning and the subprogram used, it is described below. Description of data set file: A large dataset with about one million spectral functions of the Anderson quantum impurity model. The input dataset contains the density of states (DOS) of the host material, the strength of Coulomb interaction between on-site electrons (U), and the hybridization between the host material and the impurity site (Gamma), as shown in the file input-w-4.dat. The output dataset contains the corresponding spectral functions. The continue DOS and spectral functions are stored with Chebyshev coefficients and wavelet coefficients, respectively. For the spectral functions, we save both the high-frequency and low-frequency wavelet coefficients, see the files output-w-a4 and output-w-b4. From this dataset, we build seven different machine learning networks to predict the spectral function from the input DOS, U, and Gamma. In the subprogram folder, we have cnn, resnet, gru, bigru, lstm, bilstm, bigru+gru machine learning models written in pytorch framework, as well as some auxiliary subprograms written in python, such as the data fitting programs and the plotting programs.