Impact of quantum noise on the training of quantum Generative Adversarial Networks

Abstract

Current noisy intermediate-scale quantum devices suffer from various sources of intrinsic quantum noise. Overcoming the effects of noise is a major challenge, for which different error mitigation and error correction techniques have been proposed. In this paper, we conduct a first study of the performance of quantum Generative Adversarial Networks (qGANs) in the presence of different types of quantum noise, focusing on a simplified use case in high-energy physics. In particular, we explore the effects of readout and two-qubit gate errors on the qGAN training process. Simulating a noisy quantum device classically with IBM's Qiskit framework, we examine the threshold of error rates up to which a reliable training is possible. In addition, we investigate the importance of various hyperparameters for the training process in the presence of different error rates, and we explore the impact of readout error mitigation on the results.

Conditional Born machine for Monte Carlo events generation

“…These observations suggest that the noise is assimilated during the training, underlining the importance of using actual quantum hardware. This supports the findings of Borras et al [50], which empirically found that quantum generative adversarial networks can be efficiently trained on quantum hardware if the readout noise is smaller than 0.1. Thus, QCBMs seem to be an appealing application for NISQ devices.…”

Towards a scalable discrete quantum generative adversarial neural network

“…The recent emergence of quantum machine learning [11] has lead to the proposal of various hybrid architectures, including QGANs [12][13][14][15][16][17][18][19][20][21][22]. Compared to other probabilistic models, GANs are especially well suited for implementation on quantum devices because they do not require exhaustive sampling during training.…”

Towards a scalable discrete quantum generative adversarial neural network

“…The recent emergence of quantum machine learning [11] has lead to the proposal of various hybrid architectures, including quantum generative adversarial networks (QGANs) [12][13][14][15][16][17][18][19][20][21]. Compared to other probabilistic models, GANs are especially well suited for implementation on quantum devices because they do not require exhaustive sampling during training.…”

“…As we will later see, their architectures cannot be ported directly to our binary data. Furthermore, in [13,[20][21][22] as well as to a lesser extent in [16], the authors use a classical discriminator. In [18], the authors focus on learning 1-qubit quantum states.…”

“…In [18], the authors focus on learning 1-qubit quantum states. QGANs for binary data have been studied in [14,15,20]. In these schemes the binary samples are obtained via single-shot measurements of the generator's output.…”