ShadowNet for Data-Centric Quantum System Learning

Yuxuan Du,Yibo Yang,Tongliang Liu,Zhouchen Lin,Bernard Ghanem,Dacheng Tao

2023-08-22

Abstract:Understanding the dynamics of large quantum systems is hindered by the curse of dimensionality. Statistical learning offers new possibilities in this regime by neural-network protocols and classical shadows, while both methods have limitations: the former is plagued by the predictive uncertainty and the latter lacks the generalization ability. Here we propose a data-centric learning paradigm combining the strength of these two approaches to facilitate diverse quantum system learning (QSL) tasks. Particularly, our paradigm utilizes classical shadows along with other easily obtainable information of quantum systems to create the training dataset, which is then learnt by neural networks to unveil the underlying mapping rule of the explored QSL problem. Capitalizing on the generalization power of neural networks, this paradigm can be trained offline and excel at predicting previously unseen systems at the inference stage, even with few state copies. Besides, it inherits the characteristic of classical shadows, enabling memory-efficient storage and faithful prediction. These features underscore the immense potential of the proposed data-centric approach in discovering novel and large-scale quantum systems. For concreteness, we present the instantiation of our paradigm in quantum state tomography and direct fidelity estimation tasks and conduct numerical analysis up to 60 qubits. Our work showcases the profound prospects of data-centric artificial intelligence to advance QSL in a faithful and generalizable manner.

Quantum Physics,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

The paper aims to address the challenges in Quantum System Learning (QSL), particularly the problem of characterizing large quantum systems. Specifically, the paper proposes a novel learning paradigm called ShadowNet, which combines the advantages of classical shadows methods and Neural-Network-based Quantum System Learning (NN-QSL).

The main goal of ShadowNet is to create a training dataset by utilizing classical shadows and other easily accessible information, and to use neural networks to reveal the underlying mapping rules of the explored QSL problem. This approach not only allows for efficient prediction of previously unseen systems after offline training, even with a small number of samples, but also inherits the memory efficiency and prediction accuracy advantages of classical shadows.

The paper particularly emphasizes the application instances of ShadowNet in Quantum State Tomography (QST) and Direct Fidelity Estimation (DFE) tasks. For the QST task, ShadowNet can effectively predict the density matrix of previously unseen states; for the DFE task, ShadowNet can accurately estimate the similarity, or fidelity, between two quantum states.

Overall, ShadowNet aims to overcome the limitations of existing methods in terms of generalization ability, efficiency, and prediction accuracy, thereby providing a powerful tool for exploring novel and large-scale quantum systems.