New framework extends Boltzmann generators to grand canonical ensemble

Researchers have introduced Grand Canonical Generators (GCG), a novel framework designed to extend Boltzmann generators to the grand canonical ensemble. This approach allows for the joint sampling of particle number and configuration by conditioning a variable-size generative model on chemical potential. An alternative factorized formulation leverages existing Boltzmann generators for the canonical component, analytically encoding chemical potential dependence and enabling tractable likelihoods for self-normalized importance sampling (SNIS). Empirical tests on systems like a Lennard-Jones fluid and methane adsorption in a zeolite demonstrated GCG's accuracy in reproducing grand canonical observables across varying chemical potentials. AI

IMPACT Introduces a new generative framework for simulating complex molecular systems, potentially advancing research in materials science and chemistry.

RANK_REASON The cluster contains a new academic paper detailing a novel generative framework for statistical mechanics. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• arXiv
• chemical potential
• grand canonical ensemble
• Grand Canonical Generators
• Lennard--Jones fluid
• methane
• self-normalized importance sampling
• zeolite

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →