Gauging practical computational advantage using a classical, threshold-based Gaussian boson sampler

TL;DR

This paper introduces a classical, threshold-based Gaussian boson sampler that efficiently tackles complex graph problems, demonstrating superior solutions over classical methods for large graphs with up to 2000 nodes.

Contribution

The authors present a scalable classical model of Gaussian boson sampling that effectively maps NP-Complete graph problems, showing practical advantages over traditional classical solvers.

Findings

Better solutions for dense subgraph problems in large graphs.

Efficient classical simulation of Gaussian boson sampling.

Scalable approach applicable to graphs with up to 2000 nodes.

Abstract

We describe an efficient, scalable Gaussian boson sampler based on a classical description of squeezed quantum light and a deterministic model of single-photon detectors that click when the incident amplitude falls above a given threshold. Using this model, we map several NP-Complete graph theoretic problems to equivalent Gaussian boson sampling problems and numerically explore the practical efficacy of our approach. Specifically, for a given weighted, undirected graph we examined finding the densest k-subgraph and the maximum weighted clique. We also examined the graph classification problem. Compared to traditional classical solvers, we found that our method provides better solutions in a comparable amount of samples for graphs with up to 2000 nodes.