Quantum receivers aim to effectively navigate the vast quantum-state space to endow quantum information processing capabilities unmatched by classical receivers. To date, only a handful of quantum receivers have been constructed to tackle the problem of discriminating coherent states. Quantum receivers designed by analytical approaches, however, are incapable of effectively adapting to diverse environmental conditions, resulting in their quickly diminishing performance as the operational complexities increase. Here, we present a general architecture, dubbed the quantum receiver enhanced by adaptive learning, to adapt quantum receiver structures to diverse operational conditions. The adaptively learned quantum receiver is experimentally implemented in a hardware platform with record-high efficiency. Combining the architecture and the experimental advances, the error rate is reduced up to 40% over the standard quantum limit in two coherent-state encoding schemes.

We report a field test of distribution and reconfigurable routing of multichannel entangled photons in a quantum network testbed. Nonlocal dispersion cancellation has been leveraged to maintain high-quality entanglement in the quantum network.

Adaptive quantum receiver designed by machine learning is demonstrated for discriminating multiple nonorthogonal coherent states, achieving reduced error rates of 20% (50%) over existing quantum (classical) receivers.