Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks

Abstract

Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification approaches can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined to an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers and collective variables in molecules, and can be extended to any NN potential architecture and materials system.

arXiv comment12 pages, 4 figures, supporting information

References in corpus (20)

• Escaping free-energy minima
• On representing chemical environments
• Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics
• SchNet - a deep learning architecture for molecules and materials
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• Adversarial Attacks on Neural Networks for Graph Data
• DScribe: Library of Descriptors for Machine Learning in Materials Science
• Machine Learning Unifies the Modelling of Materials and Molecules
• Towards Exact Molecular Dynamics Simulations with Machine-Learned Force Fields
• Phase transitions of hybrid perovskites simulated by machine-learning force fields trained on-the-fly with Bayesian inference
• How van der Waals interactions determine the unique properties of water
• Nucleation mechanism for the direct graphite-to-diamond phase transition
• Low-Scaling Algorithm for Nudged Elastic Band Calculations Using a Surrogate Machine Learning Model
• Committee neural network potentials control generalization errors and enable active learning
• Evidence for supercritical behavior of high-pressure liquid hydrogen
• How Neural Networks Extrapolate: From Feedforward to Graph Neural Networks
• Uncertainty estimation for molecular dynamics and sampling
• Generative Models for Automatic Chemical Design
• Automatically growing global reactive neural network potential energy surfaces: a trajectory free active learning strategy
• Challenges for Machine Learning Force Fields in Reproducing Potential Energy Surfaces of Flexible Molecules

Cited by in corpus (23)

• Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting
• Single-model uncertainty quantification in neural network potentials does not consistently outperform model ensembles
• Excited state, non-adiabatic dynamics of large photoswitchable molecules using a chemically transferable machine learning potential
• Deep Ensembles vs. Committees for Uncertainty Estimation in Neural-Network Force Fields: Comparison and Application to Active Learning
• Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials
• Machine-learning-accelerated simulations to enable automatic surface reconstruction
• Machine-learned molecular mechanics force field for the simulation of protein-ligand systems and beyond
• ArcaNN: automated enhanced sampling generation of training sets for chemically reactive machine learning interatomic potentials
• On the design space between molecular mechanics and machine learning force fields
• Model-free quantification of completeness, uncertainties, and outliers in atomistic machine learning using information theory
• Machine Learning and Data-Driven Methods in Computational Surface and Interface Science
• Data efficiency and extrapolation trends in neural network interatomic potentials
• Energy-conserving molecular dynamics is not energy conserving
• Reducing the cost of neural network potential generation for reactive molecular systems
• Four-Dimensional-Spacetime Atomistic Artificial Intelligence Models
• Prediction rigidities for data-driven chemistry
• Tell machine learning potentials what they are needed for: Simulation-oriented training exemplified for glycine
• Enhanced sampling of robust molecular datasets with uncertainty-based collective variables
• A neural-network-backed effective harmonic potential study of the ambient pressure phases of hafnia
• Learning atomic forces from uncertainty-calibrated adversarial attacks
• A Simple and Scalable Kernel Density Approach for Reliable Uncertainty Quantification in Atomistic Machine Learning
• Algorithmic differentiation for plane-wave DFT: materials design, error control and learning model parameters
• XMCQDPT2-Fidelity Transfer-Learning Potentials and a Wavepacket Oscillation Model with Power-Law Decay for Ultrafast Photodynamics