Variational Tensor Network Simulation of Gaussian Boson Sampling and Beyond

TL;DR

This paper introduces a classical variational tensor network method to simulate Gaussian Boson Sampling and other continuous variable quantum sampling problems, offering a new approach to evaluate quantum advantage.

Contribution

It reformulates continuous variable sampling as a ground state problem and employs tensor network variational methods, including novel basis optimization for non-Gaussian cases.

Findings

Achieved state-of-the-art accuracy in simulating Gaussian Boson Sampling.

Developed a new local basis optimization technique for non-Gaussian sampling.

Validated the method's effectiveness through comprehensive simulations.

Abstract

The continuous variable quantum computing platform constitutes a promising candidate for realizing quantum advantage, as exemplified in Gaussian Boson Sampling. While noise in the experiments makes the computation attainable for classical simulations, it has been suggested that the addition of non-linear elements to the experiment will help retain the quantum advantage. We propose a classical simulation tool for general continuous variable sampling problems, including Gaussian Boson Sampling and beyond. We reformulate the sampling problem as that of finding the ground state of a simple few-body Hamiltonian. This allows us to employ powerful variational methods based on tensor networks and to read off the simulation error directly from the expectation value of the Hamiltonian. We validate our method by simulating Gaussian Boson Sampling, where we achieve results comparable to the state of the art. We also consider a non-Gaussian sampling problem, for which we develop novel local basis optimization techniques based on a non-linear parameterization of the implicit basis, resulting in high effective cutoffs with diminished computational overhead.