Deep Learning Hamiltonian Monte Carlo
This work addresses sampling challenges in physics simulations, but it is incremental as it builds on existing Hamiltonian Monte Carlo methods with neural network enhancements.
The authors tackled the problem of sampling from different topologies in lattice gauge theory by generalizing Hamiltonian Monte Carlo with neural networks, resulting in successful mode mixing and significantly reduced computational cost for generating independent gauge field configurations.
We generalize the Hamiltonian Monte Carlo algorithm with a stack of neural network layers and evaluate its ability to sample from different topologies in a two dimensional lattice gauge theory. We demonstrate that our model is able to successfully mix between modes of different topologies, significantly reducing the computational cost required to generated independent gauge field configurations. Our implementation is available at https://github.com/saforem2/l2hmc-qcd .