Quantum machine learning with differential privacy

Quantum computing basicsBecause of the power of superposition and entanglement generated by quantum gates, quantum computing can create a huge speedup in certain difficult computational tasks and afford quantum advantages to ML33,34. \end{aligned} \end{aligned}$$ (6)The general single-qubit rotation can be constructed with two of the single-qubit rotations (R{x}), (R{y}), and (R_{z}). Furthermore, differential privacy gives the worst-case scenario privacy loss, thus a smaller (\varepsilon) does not necessarily mean the privacy is better. We leave the development of a tighter upper bound on privacy loss in trained quantum circuits to a future study. discuss how adding depolarizable noise to a quantum circuit imposes differential privacy on the model, providing robustness against adversarial examples76.