Are there practical uses for boson sampling?

Question

Boson sampling is a computation that is not doable on classical machines but quite doable on quantum photonic machines.

(https://www.researchgate.net/figure/Experimental-results-for-3-and-4-boson-sampling-A-B-The-equivalent-3-and-4-boson_fig2_301872712)

The precise generated pattern at the end of the board will take impossibly long to compute/simulate on classical machines.

However, just being a computation that is not doable by classical machines does not seem too rare or special... Classical computers cannot fully simulate the brain of a six-year-old for even a microsecond either.

So are there any practical uses for being able to perform boson sampling, other than doing a computation that classical machines cannot do?

Answer 1

There is a number of techniques that use Gaussian boson sampling for approximate solution of some practical problems:

https://iopscience.iop.org/article/10.1088/2058-9565/ab8504/meta

For example, some of them rely on the fact that the probability of observing a particular sample is proportional to the (squared) Hafnian of a particular matrix that specifies the GBS experiment. If the matrix is an adjacency matrix of a graph, then the Hafnian of its matrix is the number of perfect matchings in the corresponding induced subgraph. Denser subgraphs tend to have more perfect matchings. Thus, samples from GBS can be used to search for dense subgraphs in a given graph.