Experimental quantum-enhanced kernel-based machine learning on a photonic processor

This article reports the first experimental demonstration of a quantum kernel estimation using a photonic integrated processor. The approach maps data into a quantum feature space via the evolution of two-photon Fock states through a programmable photonic circuit, enabling high-accuracy binary classification. The results show that quantum kernels utilizing quantum interference outperform classical kernel methods such as Gaussian and neural tangent kernels, even with relatively small system dimensions. These findings suggest that current quantum photonic technology can enhance machine learning performance for medium-sized problems, opening pathways for practical quantum advantage in data classification tasks.