Skipping the Replica Exchange Ladder with Normalizing Flows

Abstract

We combine replica exchange (parallel tempering) with normalizing flows, a class of deep generative models. These two sampling strategies complement each other, resulting in an efficient strategy for sampling molecular systems characterized by rare events, which we call learned replica exchange (LREX). In LREX, a normalizing flow is trained to map the configurations of the fastest-mixing replica into configurations belonging to the target distribution, allowing direct exchanges between the two without the need to simulate intermediate replicas. This can significantly reduce the computational cost compared to standard replica exchange. The proposed method also offers several advantages with respect to Boltzmann generators that directly use normalizing flows to sample the target distribution. We apply LREX to some prototypical molecular dynamics systems, highlighting the improvements over previous methods.

References in corpus (6)

• Making the best of a bad situation: a multiscale approach to free energy calculation
• Flow Annealed Importance Sampling Bootstrap
• Learning Mappings between Equilibrium States of Liquid Systems Using Normalizing Flows
• Aspects of scaling and scalability for flow-based sampling of lattice QCD
• Exploration vs Convergence Speed in Adaptive-bias Enhanced Sampling
• Conditioning Normalizing Flows for Rare Event Sampling

Cited by in corpus (10)

• Enhanced Sampling in the Age of Machine Learning: Algorithms and Applications
• Roadmap on machine learning glassy dynamics
• Multimap targeted free energy estimation
• Normalizing flows as an enhanced sampling method for atomistic supercooled liquids
• Neural Thermodynamic Integration: Free Energies from Energy-based Diffusion Models
• Efficient mapping of phase diagrams with conditional Boltzmann Generators
• Scaling of Stochastic Normalizing Flows in $\mathrm{SU}(3)$ lattice gauge theory
• Solvation Free Energies from Neural Thermodynamic Integration
• Learned Mappings for Targeted Free Energy Perturbation between Peptide Conformations
• Nearest-Neighbours Neural Network architecture for efficient sampling of statistical physics models