Hierarchical modeling of molecular energies using a deep neural network

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 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.

References in corpus (4)

• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• Machine-learning based interatomic potential for amorphous carbon
• Efficient and Accurate Machine-Learning Interpolation of Atomic Energies in Compositions with Many Species
• ANI-1: A data set of 20M off-equilibrium DFT calculations for organic molecules

Cited by in corpus (86)

• Machine learning and the physical sciences
• Machine Learning Force Fields
• PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• Less is more: sampling chemical space with active learning
• Atomistic Line Graph Neural Network for Improved Materials Property Predictions
• Towards Exact Molecular Dynamics Simulations with Machine-Learned Force Fields
• Machine learning a general purpose interatomic potential for silicon
• SchNetPack: A Deep Learning Toolbox For Atomistic Systems
• Machine learning for electronically excited states of molecules
• SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effects
• sGDML: Constructing Accurate and Data Efficient Molecular Force Fields Using Machine Learning
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
• A deep potential model with long-range electrostatic interactions
• Combining SchNet and SHARC: The SchNarc machine learning approach for excited-state dynamics
• On the Completeness of Atomic Structure Representations
• Machine-learned multi-system surrogate models for materials prediction
• Machine learning of molecular properties: locality and active learning
• Feature Optimization for Atomistic Machine Learning Yields A Data-Driven Construction of the Periodic Table of the Elements
• Neural Network Potentials for Chemistry: Concepts, Applications and Prospects
• A Deep Dive into Machine Learning Density Functional Theory for Materials Science and Chemistry
• Molecular Force Fields with Gradient-Domain Machine Learning: Construction and Application to Dynamics of Small Molecules with Coupled Cluster Forces
• A Universal Density Matrix Functional from Molecular Orbital-Based Machine Learning: Transferability across Organic Molecules
• Gaussian Moments as Physically Inspired Molecular Descriptors for Accurate and Scalable Machine Learning Potentials
• Guest Editorial: Special Topic on Data-enabled Theoretical Chemistry
• End-to-End Differentiable Molecular Mechanics Force Field Construction
• Automated discovery of a robust interatomic potential for aluminum
• Strategies for the Construction of Machine-Learning Potentials for Accurate and Efficient Atomic-Scale Simulations
• Graph Neural Networks for Molecules
• Metadynamics for Training Neural Network Model Chemistries: a Competitive Assessment
• Chemical diversity in molecular orbital energy predictions with kernel ridge regression
• PiNN: A Python Library for Building Atomic Neural Networks of Molecules and Materials
• Band gap prediction for large organic crystal structures with machine learning
• Three-Dimensionally Embedded Graph Convolutional Network (3DGCN) for Molecule Interpretation
• Data-driven Material Models for Atomistic Simulation
• Regression-clustering for Improved Accuracy and Training Cost with Molecular-Orbital-Based Machine Learning
• Accelerating Finite-temperature Kohn-Sham Density Functional Theory with Deep Neural Networks
• Accurate large-scale simulations of siliceous zeolites by neural network potentials
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Artificial Intelligence in Material Engineering: A review on applications of AI in Material Engineering
• Machine learning and excited-state molecular dynamics
• Graph Neural Networks Accelerated Molecular Dynamics
• Quantum-accurate magneto-elastic predictions with classical spin-lattice dynamics
• High-Dimensional Potential Energy Surfaces for Molecular Simulations
• Molecular Force Fields with Gradient-Domain Machine Learning (GDML): Comparison and Synergies with Classical Force Fields
• Improved accuracy and transferability of molecular-orbital-based machine learning: Organics, transition-metal complexes, non-covalent interactions, and transition states
• Automated Calculation of Thermal Rate Coefficients using Ring Polymer Molecular Dynamics and Machine-Learning Interatomic Potentials with Active Learning
• Unsupervised machine learning for detection of phase transitions in off-lattice systems II. Applications
• Predicting Properties of Periodic Systems from Cluster Data: A Case Study of Liquid Water
• Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules
• Machine learning the computational cost of quantum chemistry
• Fast and Sample-Efficient Interatomic Neural Network Potentials for Molecules and Materials Based on Gaussian Moments
• Machine learning for molecular dynamics with strongly correlated electrons
• Machine Learned Hückel Theory: Interfacing Physics and Deep Neural Networks
• ArcaNN: automated enhanced sampling generation of training sets for chemically reactive machine learning interatomic potentials
• Symmetrical Graph Neural Network for Quantum Chemistry, with Dual R/K Space
• PhysNet Meets CHARMM: A Framework for Routine Machine Learning / Molecular Mechanics Simulations
• TF3P: Three-dimensional Force Fields Fingerprint Learned by Deep Capsular Network
• Accurate Molecular-Orbital-Based Machine Learning Energies via Unsupervised Clustering of Chemical Space
• Denoise Pretraining on Nonequilibrium Molecules for Accurate and Transferable Neural Potentials
• Lightweight and Effective Tensor Sensitivity for Atomistic Neural Networks
• Modeling electronic response properties with an explicit-electron machine learning potential
• Impact of the characteristics of quantum chemical databases on machine learning predictions of tautomerization energies
• Atomistic structure search using local surrogate mode
• Data efficiency and extrapolation trends in neural network interatomic potentials
• Critical Benchmarking of the G4(MP2) Model, the Correlation Consistent Composite Approach and Popular Density Functional Approximations on a Probabilistically Pruned Benchmark Dataset of Formation Enthalpies
• Size and Temperature Transferability of Direct and Local Deep Neural Networks for Atomic Forces
• Quantum chemical roots of machine-learning molecular similarity descriptors
• FeNNol: an Efficient and Flexible Library for Building Force-field-enhanced Neural Network Potentials
• Analytical Gradients for Molecular-Orbital-Based Machine Learning
• Mean-Field Density Matrix Decompositions
• Similarity Equivariant Graph Neural Networks for Homogenization of Metamaterials
• Machine Learning Models Capture Plasmon Dynamics in Ag Nanoparticles
• Construction of Machine Learned Force Fields with Quantum Chemical Accuracy: Applications and Chemical Insights
• Building nonparametric $n$-body force fields using Gaussian process regression
• Shadow molecular dynamics and atomic cluster expansions for flexible charge models
• Accurate Molecular Dynamics Enabled by Efficient Physically-Constrained Machine Learning Approaches
• Machine Learning Framework for Modeling Exciton-Polaritons in Molecular Materials
• Equivariant Tensor Network Potentials
• Reconstructing Kernel-based Machine Learning Force Fields with Super-linear Convergence
• FastCHGNet: Training one Universal Interatomic Potential to 1.5 Hours with 32 GPUs
• Machine-Learned Potentials for Solvation Modeling
• Unified theory of atom-centered representations and message-passing machine-learning schemes
• Atomistic Global Optimization X: A Python package for optimization of atomistic structures
• Knowledge Distillation of Noisy Force Labels for Improved Coarse-Grained Force Fields
• HOAX: A Hyperparameter Optimization Algorithm Explorer for Neural Networks