A new method of constructing adversarial examples of quantum variational circuits

Abstract

Abstract Quantum variational circuit is a quantum machine learning model similar to neural network. A crafted adversarial example can lead to incorrect result of the model. Using adversarial examples to train the model will greatly improve its robustness. The existing method is to use automatic differential or finite difference to get gradient and use it to construct adversarial examples. The paper proposes an innovative method to construct adversarial examples of quantum variational circuits. In the method, the gradient can be obtained by measuring the expected value of a quantum bit respectively in a series quantum circuit. This method can be used to construct the adversarial examples of a quantum variational circuit classifier. In addition, the implementation results prove the effectiveness of the proposed method. Compare with the existing method, our method requires less resources and is more efficient.