MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields

Abstract

Creating fast and accurate force fields is a long-standing challenge in computational chemistry and materials science. Recently, several equivariant message passing neural networks (MPNNs) have been shown to outperform models built using other approaches in terms of accuracy. However, most MPNNs suffer from high computational cost and poor scalability. We propose that these limitations arise because MPNNs only pass two-body messages leading to a direct relationship between the number of layers and the expressivity of the network. In this work, we introduce MACE, a new equivariant MPNN model that uses higher body order messages. In particular, we show that using four-body messages reduces the required number of message passing iterations to just two, resulting in a fast and highly parallelizable model, reaching or exceeding state-of-the-art accuracy on the rMD17, 3BPA, and AcAc benchmark tasks. We also demonstrate that using higher order messages leads to an improved steepness of the learning curves.

Advances in Neural Information Processing Systems, 2022

Cited by in corpus (47)

• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• SchNetPack 2.0: A neural network toolbox for atomistic machine learning
• MLatom 3: Platform for machine learning-enhanced computational chemistry simulations and workflows
• i-PI 3.0: a flexible and efficient framework for advanced atomistic simulations
• Accelerating CALPHAD-based Phase Diagram Predictions in Complex Alloys Using Universal Machine Learning Potentials: Opportunities and Challenges
• Deep Coarse-grained Potentials via Relative Entropy Minimization
• Lifelong Machine Learning Potentials
• Developments and Further Applications of Ephemeral Data Derived Potentials
• Advances in modeling complex materials: The rise of neuroevolution potentials
• On the design space between molecular mechanics and machine learning force fields
• Enhancing the Accuracy of Density Functional Tight Binding Models Through ChIMES Many-body Interaction Potentials
• Data efficiency and extrapolation trends in neural network interatomic potentials
• SCINE -- Software for Chemical Interaction Networks
• Stress and heat flux via automatic differentiation
• Machine-Learned Atomic Cluster Expansion Potentials for Fast and Quantum-Accurate Thermal Simulations of Wurtzite AlN
• No Headache for PIPs: A PIP Potential for Aspirin Outperforms Other Machine-Learned Potentials
• How Graph Neural Network Interatomic Potentials Extrapolate: Role of the Message-Passing Algorithm
• Probing the effects of broken symmetries in machine learning
• Uncertainty Quantification in Atomistic Simulations of Silicon using Interatomic Potentials
• A smooth basis for atomistic machine learning
• Weighted Active Space Protocol for Multireference Machine-Learned Potentials
• Iterative charge equilibration for fourth-generation high-dimensional neural network potentials
• Enhanced sampling of robust molecular datasets with uncertainty-based collective variables
• Investigating Ionic Diffusivity in Amorphous Solid Electrolytes using Machine Learned Interatomic Potentials
• EOSnet: Embedded Overlap Structures for Graph Neural Networks in Predicting Material Properties
• Scaling Graph Neural Networks to Large Proteins
• Equivariant Tensor Network Potentials
• Application-specific machine-learned interatomic potentials: exploring the trade-off between DFT convergence, MLIP expressivity, and computational cost
• Discovery of sustainable energy materials via the machine-learned material space
• Thermal Conductivity Calculation using Homogeneous Non-equilibrium Molecular Dynamics Simulation with Allegro
• Topological descriptors for the electron density of inorganic solids
• Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability
• HEroBM: a deep equivariant graph neural network for universal backmapping from coarse-grained to all-atom representations
• Free energy profiles for chemical reactions in solution from high-dimensional neural network potentials: The case of the Strecker synthesis
• Complete Neural Networks for Complete Euclidean Graphs
• Zero Shot Molecular Generation via Similarity Kernels
• Teaching oxidation states to neural networks
• Unifying O(3) Equivariant Neural Networks Design with Tensor-Network Formalism
• Implicit Neural Representations for Chemical Reaction Paths
• Elucidating Ion Capture and Transport Mechanisms of Preyssler Anions in Aqueous Solutions Using Biased MACE-Accelerated MD Simulations
• Systematic Study of Machine Learning Classification Algorithms of Zeolitic Imidazolate Framework Polymorphs
• Machine Learning Photodynamics Unveils a Controlled H${}_2$ Loss Channel in Methaniminium Cation
• A Simple and Scalable Kernel Density Approach for Reliable Uncertainty Quantification in Atomistic Machine Learning
• Exact average many-body interatomic interaction model for random alloys
• XMCQDPT2-Fidelity Transfer-Learning Potentials and a Wavepacket Oscillation Model with Power-Law Decay for Ultrafast Photodynamics
• Knowledge Distillation of a Protein Language Model Yields a Foundational Implicit Solvent Model
• Neural Polarization: Toward Electron Density for Molecules by Extending Equivariant Networks