Quantum computers promise to shift the paradigm for high-performance computing in the years to come. While large-scale, fault-tolerant machines capable of universal quantum computation are a distant goal, devices that can harness quantum-mechanical effects to solve a specific task faster than a classical computer are within reach. Here, we build an optical setup that implements an example of such a device—a Gaussian boson sampling (GBS) machine—dedicated to finding dense subgraphs of a graph, a task relevant to many questions in network science, graph compression, and biology. One could use it, for example, to find dense communities in social networks.

A GBS machine works by injecting nonclassical light into an optical circuit and measuring where the photons come out using single-photon detectors. By choosing the appropriate parameters for the input light and interferometer, GBS devices can be used to identify complex features of graphs or calculate the molecular vibronic spectra of molecules, among other applications. Even the largest supercomputers struggle to find solutions for such tasks.

In our work, we demonstrate a speedup in finding subgraphs when using the outputs from the GBS device compared with using random samples in a classical algorithm. Moreover, we use the temporal degree of freedom to encode the information in a novel way, opening a new door to solve scalability issues for future experiments.

We foresee the scaling up of similar approaches to apply GBS devices for practical problems. Our work provides a proof-of-principle example that the nonuniversal, but practical, approach to quantum computing is an avenue to solving dedicated problems with far fewer resources compared to a universal quantum computer.