Generative Flow Networks in Covariant Loop Quantum Gravity

TL;DR

This paper introduces the use of Generative Flow Networks, a machine learning technique, to compute geometric observables in covariant loop quantum gravity, demonstrating its effectiveness compared to traditional stochastic algorithms.

Contribution

It applies Generative Flow Networks to spin foam models in quantum gravity, providing a novel computational approach for evaluating geometric quantities.

Findings

GFlowNets successfully compute dihedral angles in 4-simplices.

Results are consistent with previous MCMC-based methods.

Demonstrates potential for scalable quantum gravity simulations.

Abstract

Spin foams arose as the covariant (path integral) formulation of quantum gravity depicting transition amplitudes between different quantum geometry states. As such, they provide a scheme to study the no boundary proposal, specifically the nothing to something transition and compute relevant observables using high performance computing (HPC). Following recent advances, where stochastic algorithms (Markov Chain Monte Carlo-MCMC) were used, we employ Generative Flow Networks, a newly developed machine learning algorithm to compute the expectation value of the dihedral angle for a 4-simplex and compare the results with previous works.