fig11

Figure 11. The general process of training and applying MLPs. Reprinted with permission^[76]. Copyright 2024, Elsevier. MLPs: Machine learning potentials; AIMD: Ab initio molecular dynamics; SQS: Special quasirandom structures; ASCF: ab initio self-consistent field; wASCF: weighted atom-centered symmetry functions; MTP: moment tensor potential; SOAP: smooth overlap of atomic positions; ACE: atomic cluster expansion; SNAP: spectral neighbor analysis potentials; DNN: deep neural network; RNN: recurrent neural network; NEP: neuroevolution potential; CNN: convolutional neural network; GCNN: graph convolutional neural network; EGNN: equivariant graph neural network; NequlP: neural equivariant interatomic potential; GAP: gaussian approximation potential; MLIP: machine learning interatomic potential.