Learning Pair Potentials using Differentiable Simulations

Abstract

Learning pair interactions from experimental or simulation data is of great interest for molecular simulations. We propose a general stochastic method for learning pair interactions from data using differentiable simulations (DiffSim). DiffSim defines a loss function based on structural observables, such as the radial distribution function, through molecular dynamics (MD) simulations. The interaction potentials are then learned directly by stochastic gradient descent, using backpropagation to calculate the gradient of the structural loss metric with respect to the interaction potential through the MD simulation. This gradient-based method is flexible and can be configured to simulate and optimize multiple systems simultaneously. For example, it is possible to simultaneously learn potentials for different temperatures or for different compositions. We demonstrate the approach by recovering simple pair potentials, such as Lennard-Jones systems, from radial distribution functions. We find that DiffSim can be used to probe a wider functional space of pair potentials compared to traditional methods like Iterative Boltzmann Inversion. We show that our methods can be used to simultaneously fit potentials for simulations at different compositions and temperatures to improve the transferability of the learned potentials.

arXiv comment12 pages, 10 figures

References in corpus (21)

• Adam: A Method for Stochastic Optimization
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Solving high-dimensional partial differential equations using deep learning
• Machine Learning Force Fields
• Interatomic potentials from first-principles calculations: the force-matching method
• Strong scaling of general-purpose molecular dynamics simulations on GPUs
• A Model of Inductive Bias Learning
• Neural Ordinary Differential Equations
• Representability problems for coarse-grained water potentials
• Learning Deep Embeddings with Histogram Loss
• TorchMD: A deep learning framework for molecular simulations
• Differentiable Programming Tensor Networks
• Derivation of coarse-grained potentials via multistate iterative Boltzmann inversion
• DeePCG: constructing coarse-grained models via deep neural networks
• Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting
• Learning the exchange-correlation functional from nature with fully differentiable density functional theory
• Automatic differentiation in quantum chemistry with an application to fully variational Hartree-Fock
• Efficient and Modular Implicit Differentiation
• SPNets: Differentiable Fluid Dynamics for Deep Neural Networks
• Differentiable Molecular Simulations for Control and Learning

Cited by in corpus (6)

• Accurate machine learning force fields via experimental and simulation data fusion
• PySAGES: flexible, advanced sampling methods accelerated with GPUs
• Spiers Memorial Lecture: How to do impactful research in artificial intelligence for chemistry and materials science
• Implicit Chain Particle Model for Polymer Grafted Nanoparticles
• Metadensity functional theory for classical fluids: Extracting the pair potential
• Data-driven discovery of dynamics from time-resolved coherent scattering