Supervised Learning Enhanced by an Entangled Sensor Network

Various existing quantum supervised learning (SL) schemes rely on quantum random access memories to store quantum-encoded data given a priori in a classical description. The data acquisition process, however, has not been accounted for, while it sets the ultimate limit of the usefulness of the data for different SL tasks, as constrained by the quantum Cramer-Rao bound. We introduce supervised learning enhanced by an entangled sensor network (SLEEN) as a means to carry out SL tasks at the physical layer where a quantum advantage is achieved. The entanglement shared by different sensors boosts the performance of extracting global features of the object under investigation. We leverage SLEEN to construct an entanglement-enhanced support-vector machine for quantum data classification and entanglement-enhanced principal component analyzer for quantum data compression. In both schemes, variational circuits are employed to seek the optimum entangled probe state and measurement settings to maximize the entanglement-enabled quantum advantage. We compare the performance of SLEEN with separable-state SL schemes and observe an appreciable entanglement-enabled performance gain even in the presence of loss. SLEEN is realizable with available technology, opening a viable route toward building near-term quantum devices that offer unmatched performance beyond what the optimum classical device is able to afford.