Solving Quantum Statistical Mechanics with Variational Autoregressive Networks and Quantum Circuits

Setup

Typing ] in a Julia REPL, and then

pkg> dev https://github.com/wangleiphy/BetaVQE.jl.git

To make sure BetaVQE is installed properly, type

pkg> test BetaVQE

Run

Run this to train the transverse field Ising model, open a terminal and type

$ cd ~/.julia/dev/BetaVQE

$ julia --project runner.jl learn 2 2 2.0 2.0

For Windows operation system, the Julia develop folder might be different.

This utility accepts the following arguments

• nx::Int=2, lattice size in x direction,
• ny::Int=2, lattice size in y direction,
• Γ::Real=1.0, the strength of transverse field,
• β::Real=1.0, inverse temperature,

and keyword arguments

• depth::Int=5, circuit depth,
• nsamples::Int=1000, the batch size used in training,
• nhiddens::Vector{Int}=[500], dimension of the VAN's hidden layer,
• lr::Real=0.01, the learning rate of the ADAM optimizer,
• niter::Int=500, number of iteration,
• cont::Bool=false, continue from checkpoint if true.

Paper

arXiv:1912.11381