Fig. 8: General schematic diagram of DNN-CE models and the workflow of transfer learning method in this work.

a DNN-CE includes an input layer of descriptor (corresponding to all descriptors generated by CE feature model in this study), several hidden layers and an output layer (predicted formation energy). Each circle represents a neuron. Each circle represents a neuron. b A single neuron as the calculation unit of the neural network. The neuron receives the input signal through a weighted connection and performs a weighted summation on it, and performs a non-linear transformation of the calculation result through the activation function to produce an output value. c The general workflow of predicting the formation energy of perovskite oxides through the transfer learning method.