SchNet: A continuous-filter convolutional neural network for modeling quantum interactions

Abstract

Deep learning has the potential to revolutionize quantum chemistry as it is ideally suited to learn representations for structured data and speed up the exploration of chemical space. While convolutional neural networks have proven to be the first choice for images, audio and video data, the atoms in molecules are not restricted to a grid. Instead, their precise locations contain essential physical information, that would get lost if discretized. Thus, we propose to use continuous-filter convolutional layers to be able to model local correlations without requiring the data to lie on a grid. We apply those layers in SchNet: a novel deep learning architecture modeling quantum interactions in molecules. We obtain a joint model for the total energy and interatomic forces that follows fundamental quantum-chemical principles. This includes rotationally invariant energy predictions and a smooth, differentiable potential energy surface. Our architecture achieves state-of-the-art performance for benchmarks of equilibrium molecules and molecular dynamics trajectories. Finally, we introduce a more challenging benchmark with chemical and structural variations that suggests the path for further work.

References in corpus (11)

• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Quantum-Chemical Insights from Deep Tensor Neural Networks
• Deep Convolutional Networks on Graph-Structured Data
• Machine learning prediction errors better than DFT accuracy
• Finding Density Functionals with Machine Learning
• How to represent crystal structures for machine learning: towards fast prediction of electronic properties
• Xception: Deep Learning with Depthwise Separable Convolutions
• A method for the estimation of the significance of cross-correlations in unevenly sampled red-noise time series

Cited by in corpus (84)

• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• A Universal Graph Deep Learning Interatomic Potential for the Periodic Table
• Towards Exact Molecular Dynamics Simulations with Machine-Learned Force Fields
• ChemRL-GEM: Geometry Enhanced Molecular Representation Learning for Property Prediction
• Deep Parametric Continuous Convolutional Neural Networks
• Graph Representation Learning in Biomedicine
• Autonomous discovery in the chemical sciences part II: Outlook
• Higher-Order Explanations of Graph Neural Networks via Relevant Walks
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
• Solving Many-Electron Schrödinger Equation Using Deep Neural Networks
• Perspective on integrating machine learning into computational chemistry and materials science
• Inverse design of 3d molecular structures with conditional generative neural networks
• Coarse Graining Molecular Dynamics with Graph Neural Networks
• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• Representations of molecules and materials for interpolation of quantum-mechanical simulations via machine learning
• DPA-2: a large atomic model as a multi-task learner
• Machine Learning on Neutron and X-Ray Scattering
• Deep convolutional networks for quality assessment of protein folds
• AdsorbML: A Leap in Efficiency for Adsorption Energy Calculations using Generalizable Machine Learning Potentials
• Message-passing neural networks for high-throughput polymer screening
• Toward Explainable AI for Regression Models
• TeaNet: universal neural network interatomic potential inspired by iterative electronic relaxations
• Informing Geometric Deep Learning with Electronic Interactions to Accelerate Quantum Chemistry
• Second-Order Pooling for Graph Neural Networks
• SchNetPack 2.0: A neural network toolbox for atomistic machine learning
• Theoretical Guarantees for Permutation-Equivariant Quantum Neural Networks
• Solid Harmonic Wavelet Scattering for Predictions of Molecule Properties
• Fast Uncertainty Estimates in Deep Learning Interatomic Potentials
• Hierarchical, rotation-equivariant neural networks to select structural models of protein complexes
• Bayesian Force Fields from Active Learning for Simulation of Inter-Dimensional Transformation of Stanene
• Deep Ensembles vs. Committees for Uncertainty Estimation in Neural-Network Force Fields: Comparison and Application to Active Learning
• GraphVAMPNet, using graph neural networks and variational approach to markov processes for dynamical modeling of biomolecules
• Improved accuracy and transferability of molecular-orbital-based machine learning: Organics, transition-metal complexes, non-covalent interactions, and transition states
• Dynamical Strengthening of Covalent and Non-Covalent Molecular Interactions by Nuclear Quantum Effects at Finite Temperature
• Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials
• Deep Coarse-grained Potentials via Relative Entropy Minimization
• Accurate machine learning force fields via experimental and simulation data fusion
• A deep neural network for molecular wave functions in quasi-atomic minimal basis representation
• Towards Semi-supervised Universal Graph Classification
• Automated Calculation of Thermal Rate Coefficients using Ring Polymer Molecular Dynamics and Machine-Learning Interatomic Potentials with Active Learning
• Predicting Properties of Periodic Systems from Cluster Data: A Case Study of Liquid Water
• A Tungsten Deep Neural-Network Potential for Simulating Mechanical Property Degradation Under Fusion Service Environment
• Drug-Target Interaction/Affinity Prediction: Deep Learning Models and Advances Review
• E(n)-Equivariant Cartesian Tensor Passing Potential
• On the design space between molecular mechanics and machine learning force fields
• Piezoelectric modulus prediction using machine learning and graph neural networks
• SpreadGNN: Serverless Multi-task Federated Learning for Graph Neural Networks
• Machine Learning and Data-Driven Methods in Computational Surface and Interface Science
• Convolution, aggregation and attention based deep neural networks for accelerating simulations in mechanics
• Fast and flexible long-range models for atomistic machine learning
• Geometry-enhanced Pre-training on Interatomic Potentials
• Towards Graph Self-Supervised Learning with Contrastive Adjusted Zooming
• CREMP: Conformer-rotamer ensembles of macrocyclic peptides for machine learning
• Stress and heat flux via automatic differentiation
• Reducing Training Data Needs with Minimal Multilevel Machine Learning (M3L)
• Optimal pre-train/fine-tune strategies for accurate material property predictions
• Molecular Quantum Chemical Data Sets and Databases for Machine Learning Potentials
• BatmanNet: Bi-branch Masked Graph Transformer Autoencoder for Molecular Representation
• Wavelet Scattering Networks for Atomistic Systems with Extrapolation of Material Properties
• Mean-Field Density Matrix Decompositions
• Similarity Equivariant Graph Neural Networks for Homogenization of Metamaterials
• Beyond MD17: the reactive xxMD dataset
• Exact conservation laws for neural network integrators of dynamical systems
• Learning a Continuous Representation of 3D Molecular Structures with Deep Generative Models
• Towards Symbolic XAI -- Explanation Through Human Understandable Logical Relationships Between Features
• Fitting to magnetic forces improves the reliability of magnetic Moment Tensor Potentials
• Beyond Euclid: An Illustrated Guide to Modern Machine Learning with Geometric, Topological, and Algebraic Structures
• Discovery of structure-property relations for molecules via hypothesis-driven active learning over the chemical space
• Developing Potential Energy Surfaces for Graphene-based 2D-3D Interfaces from Modified High Dimensional Neural Networks for Applications in Energy Storage
• Molecular Dipole Moment Learning via Rotationally Equivariant Gaussian Process Regression with Derivatives in Molecular-orbital-based Machine Learning
• Molecular relaxation by reverse diffusion with time step prediction
• Scaling Molecular Dynamics with ab initio Accuracy to 149 Nanoseconds per Day
• Equivariant Tensor Network Potentials
• Molecular distance matrix prediction based on graph convolutional networks
• Complete Neural Networks for Complete Euclidean Graphs
• HEroBM: a deep equivariant graph neural network for universal backmapping from coarse-grained to all-atom representations
• FastCHGNet: Training one Universal Interatomic Potential to 1.5 Hours with 32 GPUs
• Learning physics confers pose-sensitivity in structure-based virtual screening
• Considerations in the use of ML interaction potentials for free energy calculations
• GeoT: A Geometry-aware Transformer for Reliable Molecular Property Prediction and Chemically Interpretable Representation Learning
• Scalable Training of Neural Network Potentials for Complex Interfaces Through Data Augmentation
• Improving Molecular Force Fields Across Configurational Space by Combining Supervised and Unsupervised Machine Learning
• Neural Mode Jump Monte Carlo
• On the forbidden graphene's ZO (out-of-plane optic) phononic band-analog vibrational modes in fullerenes