Based on the linearity of quantum unitary operations, we propose a method that runs the parameterized quantum circuits before encoding the input data. It enables a dataset owner to train machine learning models on quantum cloud computation platforms, without the risk of leaking the information of the data. It is also capable of encoding a huge number of data effectively at a later time using classical computations, thus saving the runtime on quantum computation devices. The trained quantum machine learning model can be run completely on classical computers, so that the dataset owner does not need to have any quantum hardware, nor even quantum simulators. Moreover, the method can mitigate the encoding bottom neck by reducing the required circuit depth from $O(2^{n})$ to $n/2$. These results manifest yet another advantage of quantum and quantum-inspired machine learning models over existing classical neural networks, and broaden the approaches for data security.

本研究解决了数据所有者在使用量子云计算平台训练机器学习模型时，数据安全性的问题。提出了一种在编码输入数据之前运行参数化量子电路的方法，显著降低了所需电路深度，并允许训练后的模型在经典计算机上运行，从而拓展了数据安全的解决方案。此方法为量子与量子灵感的机器学习模型提供了另一个优势，并提高了数据保护的可行性。

在不上传数据的情况下，云端训练量子机器学习模型