How to enhance quantum generative adversarial learning of noisy information

Abstract

Quantum machine learning is where nowadays machine learning (ML) meets quantum information science. In order to implement this new paradigm for novel quantum technologies, we still need a much deeper understanding of its underlying mechanisms, before proposing new algorithms to feasibly address real problems. In this context, quantum generative adversarial learning is a promising strategy to use quantum devices for quantum estimation or generative ML tasks. However, the convergence behaviours of its training process, which is crucial for its practical implementation on quantum processors, have not been investigated in detail yet. Indeed here we show how different training problems may occur during the optimization process, such as the emergence of limit cycles. The latter may remarkably extend the convergence time in the scenario of mixed quantum states playing a crucial role in the already available noisy intermediate scale quantum devices. Then, we propose new strategies to achieve a faster convergence in any operating regime. Our results pave the way for new experimental demonstrations of such hybrid classical-quantum protocols allowing to evaluate the potential advantages over their classical counterparts.

A quantum generative adversarial network for distributions

“…4). This approach of building the circuit is new since in the papers that use quantum discriminators, the circuits that are used are what is called ansatz circuits (Braccia et al 2021), in other words generic circuits built with layers of rotation gates and controlled rotation gates (see 3.6 and 3.7 below for the definition of these gates). Such ansatz circuits are therefore parameterised circuits as put forward in Chakrabarti et al (2019), where generally an interpretation on the circuit's architecture performing as a classifying neural network cannot be made.…”

“…Such ansatz circuits are therefore parameterised circuits as put forward in Chakrabarti et al (2019), where generally an interpretation on the circuit's architecture performing as a classifying neural network cannot be made. As pointed out in Braccia et al (2021), the architectures of both the generator and the discriminator are the same, which on the one hand solves the issue of having to monitor whether there is a imbalance in terms of expressivity between the generator and the discriminator; however, on the other hand,3 it prevents us from being able to give a straightforward interpretation for the given architectures. The main task here is then to translate these classical computations to a quantum input for the discriminator.…”

Quantum Quantile Mechanics: Solving Stochastic Differential Equations for Generating Time‐Series

“…This corresponds to a quantum circuit Born machine (QCBM) generator. [56] In this case qGAN was applied to sampling from discrete distributions, [57][58][59][60][61][62][63] preparing arbitrary quantum states, [64,65] and being demonstrated experimentally with superconducting circuits for image generation. [66] In this setting the sampling procedure is efficient, but its power depends on the register width and the generator is difficult to train at increasing scale.…”

Quantum Quantile Mechanics: Solving Stochastic Differential Equations for Generating Time-Series

“…This corresponds to a quantum circuit Born machine (QCBM) generator [56]. In this case qGAN was applied to sampling from discrete distributions [57][58][59][60][61][62][63], preparing arbitrary quantum states [64,65], and being demonstrated experimentally with superconducting circuits for image generation [66]. In this setting the sampling procedure is efficient, but its power depends on the register width and the generator is difficult to train at increasing scale.…”