Hierarchical modeling of molecular energies using a deep neural network

2017/09/29 by Nicholas Lubbers, Justin S. Smith, Kipton Barros · 300 citations
Materials Science · Mathematics · Physics and Astronomy · #Advanced Chemical Physics Studies #Artificial neural network #Deep neural networks #Energy (signal processing) #Machine Learning in Materials Science #Nonlinear system #Particle (ecology) #Quantum #Quantum many-body systems #Representation (politics) #State (computer science) #physics.chem-ph #stat.ML

paper · pdf · doi:10.1063/1.5011181

published in The Journal of Chemical Physics 148(24), 241715 (American Institute of Physics)

arxiv created 2017/09/29 · openalex created_date 2017/10/20 · openalex publication_date 2018/03/19 · arxiv updated 2018/04/04 · openalex updated_date 2026/08/05

Abstract

We introduce the Hierarchically Interacting Particle Neural Network (HIP-NN) to model molecular properties from datasets of quantum calculations. Inspired by a many-body expansion, HIP-NN decomposes properties, such as energy, as a sum over hierarchical terms. These terms are generated from a neural network-a composition of many nonlinear transformations-acting on a representation of the molecule. HIP-NN achieves the state-of-the-art performance on a dataset of 131k ground state organic molecules and predicts energies with 0.26 kcal/mol mean absolute error. With minimal tuning, our model is also competitive on a dataset of molecular dynamics trajectories. In addition to enabling accurate energy predictions, the hierarchical structure of HIP-NN helps to identify regions of model uncertainty.

Citations

• Machine Learning for Quantum Mechanical Properties of Atoms in Molecules
• Gaussian approximation potentials: A brief tutorial introduction
• Generalized Gradient Approximation Made Simple
• Constructing high‐dimensional neural network potentials: A tutorial review
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• Self-Consistent Equations Including Exchange and Correlation Effects
• Molecular Graph Convolutions: Moving Beyond Fingerprints
• Quantum-chemical insights from deep tensor neural networks
• Deep learning

Cited by

• Construction of Machine Learned Force Fields with Quantum Chemical Accuracy: Applications and Chemical Insights
• Symmetrical Graph Neural Network for Quantum Chemistry with Dual Real and Momenta Space