Inversion of the chemical environment representations

Abstract

Machine-learning generative methods for material design are constructed by representing a given chemical structure, either a solid or a molecule, over appropriate atomic features, generally called structural descriptors. These must be fully descriptive of the system, must facilitate the training process and must be invertible, so that one can extract the atomic configurations corresponding to the output of the model. In general, this last requirement is not automatically satisfied by the most efficient structural descriptors, namely the representation is not directly invertible. Such drawback severely limits our freedom of choice in selecting the most appropriate descriptors for the problem, and thus our flexibility to construct generative models. In this work, we present a general optimization method capable of inverting any local many-body descriptor of the chemical environment, back to a cartesian representation. The algorithm is then implemented together with the bispectrum representation of the local structure and demonstrated for a number of molecules. The scheme presented here, thus, represents a general approach to the inversion of structural descriptors, enabling the construction of efficient structural generative models.

References in corpus (5)

• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• A Spectral Analysis Method for Automated Generation of Quantum-Accurate Interatomic Potentials
• Physics-inspired structural representations for molecules and materials
• High-throughput discovery of novel cubic crystal materials using deep generative neural networks
• Through the eyes of a descriptor: Constructing complete, invertible descriptions of atomic environments

Cited by in corpus (2)

• Linear Jacobi-Legendre expansion of the charge density for machine learning-accelerated electronic structure calculations
• Zero Shot Molecular Generation via Similarity Kernels