Normalizing flows as an enhanced sampling method for atomistic supercooled liquids

Abstract

Normalizing flows can transform a simple prior probability distribution into a more complex target distribution. Here, we evaluate the ability and efficiency of generative machine learning methods to sample the Boltzmann distribution of an atomistic model for glass-forming liquids. This is a notoriously difficult task, as it amounts to ergodically exploring the complex free energy landscape of a disordered and frustrated many-body system. We optimize a normalizing flow model to successfully transform high-temperature configurations of a dense liquid into low-temperature ones, near the glass transition. We perform a detailed comparative analysis with established enhanced sampling techniques developed in the physics literature to assess and rank the performance of normalizing flows against state-of-the-art algorithms. We demonstrate that machine learning methods are very promising, showing a large speedup over conventional molecular dynamics. Normalizing flows show performances comparable to parallel tempering and population annealing, while still falling far behind the swap Monte Carlo algorithm. Our study highlights the potential of generative machine learning models in scientific computing for complex systems, but also points to some of its current limitations and the need for further improvement.

References in corpus (15)

• Generative Adversarial Networks
• Theoretical perspective on the glass transition and amorphous materials
• GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation
• Stochastic Normalizing Flows
• Towards Predicting Equilibrium Distributions for Molecular Systems with Deep Learning
• Sampling efficiency of transverse forces in dense liquids
• Sampling with flows, diffusion and autoregressive neural networks: A spin-glass perspective
• Irreversible Monte Carlo algorithms for hard disk glasses: from event-chain to collective swaps
• Dynamic heterogeneity at the experimental glass transition predicted by transferable machine learning
• Single-parameter aging in a binary Lennard-Jones system
• A Boltzmann generator for the isobaric-isothermal ensemble
• Learning Mappings between Equilibrium States of Liquid Systems Using Normalizing Flows
• Designing losses for data-free training of normalizing flows on Boltzmann distributions
• SE(3) Equivariant Augmented Coupling Flows
• Measuring glass entropies with population annealing

Cited by in corpus (9)

• Roadmap on machine learning glassy dynamics
• Monte Carlo simulations of glass-forming liquids beyond Metropolis
• Policy-guided Monte Carlo on general state spaces: Application to glass-forming mixtures
• Graph neural network-based structural classification of glass-forming liquids and its interpretation via self-attention mechanism
• Characterising the slow dynamics of the swap Monte Carlo algorithm
• Irreversible swap algorithms for soft sphere glasses
• Molecular motion at the experimental glass transition
• Computational Methods toward Ultrastable Glasses
• Sampling the Liquid-Gas Critical Point with Boltzmann Generators