Quantum Noise Sensing by Generating Fake Noise

Noisy intermediate-scale quantum (NISQ) devices are nowadays starting to become available the final user, hence potentially allowing speedups predicted by quantum-information theory be shown. However, before implementing any algorithm, it is crucial have at least a partial or possibly full knowledge on type and amount of noise affecting machine. Here, generalizing generative adversarial learning from states (QGANs) operations, superoperators, channels (here named super QGANs), we propose very promising framework characterize in realistic device, even case spatially temporally correlated (memory channels) circuits. The key idea learn about mimicking way that one cannot distinguish between real (to sensed) fake (generated) one. We find that, when applied benchmarking Pauli channels, super-QGAN protocol able associated error rates noise. Moreover, also show how employ for metrology applications. believe our QGANs pave development hybrid quantum-classical machine-learning protocols better characterization control current future unavoidably noisy devices.