fig5

Figure 5. (A) Visualization of the local environment representations learned from the elemental boron dataset. The color of each plot is coded with learned local energy. Reproduced with permission^[71]. Copyright 2018, AIP Publishing; (B) A diagram of the autoencoder used for molecular design, including the joint property prediction model; (C) Different types of molecular representations applied to one molecule. Reproduced with permission^[80]. Copyright 2018, American Association for the Advancement of Science; (D) A scheme of how TS-MGCN works; (E) Model Structure of molecular distance matrix prediction mode. Reproduced with permission^[81]. Copyright 2022, Elsevier; (F) Overview of a MEGNet module. Reproduced with permission^[41]. Copyright 2019, American Chemical Society. MEGNet: matErials graph network; MLP: multilayer perceptron; SMILES: simplified molecular input line entry specification.