Experimental benchmark of boson sampling with pattern recognition techniques.
Agresti I., Viggianiello N., Flamini F., Spagnolo N., Crespi A., Osellame R., Wiebe N., Sciarrino F.
Validating large-scale quantum devices, like boson samplers, is a major challenge to show quantum advantages over classical hardware. We propose a novel data-driven approach based on unsupervised machine learning algorithms. We train a classifier that uses $K$-means clustering to distinguish boson samplers that use indistinguishable photons and those that do not. We train the model on numerical simulations of small-scale boson samplers and validate the pattern recognition technique on larger numerical simulations and on integrated photonic platforms in boson sampling and scattershot experiments. This approach performs substantially better than previous methods and underlines the ability to further generalize its operation.