e3nn: Euclidean Neural Networks

2022/07/18 by Mario Geiger, Tess Smidt, Geiger, Mario +1 · 47 citations
Computer Science · Engineering · #3D Shape Modeling and Analysis #Medical Imaging and Analysis #Neural Networks and Applications #cs.AI #cs.LG #cs.NE

paper · pdf · doi:10.48550/arxiv.2207.09453

draft

arxiv created 2022/07/18 · arxiv updated 2022/07/21

Abstract

We present e3nn, a generalized framework for creating E(3) equivariant trainable functions, also known as Euclidean neural networks. e3nn naturally operates on geometry and geometric tensors that describe systems in 3D and transform predictably under a change of coordinate system. The core of e3nn are equivariant operations such as the TensorProduct class or the spherical harmonics functions that can be composed to create more complex modules such as convolutions and attention mechanisms. These core operations of e3nn can be used to efficiently articulate Tensor Field Networks, 3D Steerable CNNs, Clebsch-Gordan Networks, SE(3) Transformers and other E(3) equivariant networks.

Cited by

• A Combined Tight Binding with Machine Learning Potential Model for Magnesium Compounds
• DiffeoMorph: Learning to Morph 3D Shapes Using Differentiable Agent-Based Simulations
• A Single Architecture for Representing Invariance Under Any Space Group
• Equivariant symmetry-aware head pose estimation for fetal MRI
• Predicting the Thermal Behavior of Semiconductor Defects with Equivariant Neural Networks
• GLOBE: Accurate and Generalizable PDE Surrogates using Domain-Inspired Architectures and Equivariances
• Group-Theoretic Structure Governing Identifiability in Inverse Problems
• Field theoretic atomistics: Learning thermodynamic and variational surrogate to density functional theory
• Fine-Tuning Unifies Foundational Machine-learned Interatomic Potential Architectures at ab initio Accuracy
• Transformer Atomic Cluster Expansion: TRACE
• Graph Neural Network for Unified Electronic and Interatomic Potentials: Strain-tunable Electronic Structures in 2D Materials
• Equivariant Neural Networks for General Linear Symmetries on Lie Algebras
• HollowFlow: Efficient Sample Likelihood Evaluation using Hollow Message Passing
• Universally Invariant Learning in Equivariant GNNs
• Optimizing Cross-Domain Transfer for Universal Machine Learning Interatomic Potentials
• CoeffNet : predicting activation barriers through a chemically-interpretable, equivariant and physically constrained graph neural network
• Towards A Transferable Acceleration Method for Density Functional Theory
• Adaptive Canonicalization with Application to Invariant Anisotropic Geometric Networks
• Shoot from the HIP: Hessian Interatomic Potentials without derivatives
• Advancing Universal Deep Learning for Electronic-Structure Hamiltonian Prediction of Materials
• Automated Discovery of Algorithms for Molecular Electronic Structure Calculations Using Physics-Informed Program Synthesis
• Carbohydrate NMR chemical shift prediction by GeqShift employing E(3) equivariant graph neural networks
• Accelerating Protein Molecular Dynamics Simulation with DeepJump
• Energy-Based Flow Matching for Generating 3D Molecular Structure
• Equivariant Electronic Hamiltonian Prediction with Many-Body Message Passing
• Constructing 3D Rotational Invariance and Equivariance with Symmetric Tensor Networks
• Multi-head committees enable direct uncertainty prediction for atomistic foundation models
• Hybrid Quantum--Classical Machine Learning Potential with Variational Quantum Circuits
• Enhancing Materials Discovery with Valence Constrained Design in Generative Modeling
• PIL-Net: a physics-informed graph convolutional network for predicting atomic multipoles
• DualEquiNet: A Dual-Space Hierarchical Equivariant Network for Large Biomolecules
• TAGTorch: A PyTorch Library for Geometry, Topology, and Symmetry-Aware Machine Learning
• SE(3)-Equivariant Diffusion Policy in Spherical Fourier Space
• A Study on the Fine-Tuning Performance of Universal Machine-Learned Interatomic Potentials (U-MLIPs)
• The Price of Freedom: Exploring Expressivity and Runtime Tradeoffs in Equivariant Tensor Products
• GEARS H: Accurate machine-learned Hamiltonians for next-generation device-scale modeling
• Accurate Prediction of Tensorial Spectra Using Equivariant Graph Neural Network
• A Tale of Two Symmetries: Exploring the Loss Landscape of Equivariant Models
• EquAct: An SE(3)-Equivariant Multi-Task Transformer for Open-Loop Robotic Manipulation
• High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction
• Equivariant Flow Matching for Point Cloud Assembly
• Local-Global Associative Frames for Symmetry-Preserving Crystal Structure Modeling
• Probing the Temporal Response of Liquid Water to a THz Pump Pulse Using Machine Learning-Accelerated Non-Equilibrium Molecular Dynamics
• InvDesFlow-AL: active learning-based workflow for inverse design of functional materials
• Graph Neural Network Predictions of Carbon 1s Binding Energies with Near-Experimental Accuracy
• Energy-Based Coarse-Graining in Molecular Dynamics: A Flow-Based Framework without Data
• Crystal structure prediction with host-guided inpainting generation and foundation potentials
• Data-Driven Molecular Dynamics and TEM Analysis of Crystal Growth and Hydrogen Sensing in Pt-Functionalized Graphene Chemiresistive Sensors

Related

• E3x: E(3)-Equivariant Deep Learning Made Easy
• Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
• Geometric and Physical Quantities Improve E(3) Equivariant Message Passing
• A Complete Guide to Spherical Equivariant Graph Transformers
• Relaxed Equivariant Graph Neural Networks