https://doi.org/10.1140/epjp/s13360-023-04537-6
Regular Article
A progressive predictor-based quantum architecture search with active learning
1
School of Electronic and Information Engineering, Foshan University, 528000, Foshan, China
2
School of Mechatronic Engineering and Automation, Foshan University, 528000, Foshan, China
3
Peng Cheng Laboratory, 518055, Shenzhen, China
4
College of Computer and Information Engineering, Henan Normal University, 453007, Xinxiang, China
5
College of Mathematics and Informatics, South China Agricultural University, 510642, Guangzhou, China
Received:
22
July
2023
Accepted:
29
September
2023
Published online:
10
October
2023
Quantum architecture search (QAS) algorithms have demonstrated remarkable proficiency in automatically designing high-performance quantum circuits for variational quantum algorithms. Predictor-based QAS (PQAS) is an effective technique to accelerate QAS. It estimates the approximate performances of quantum circuits using a neural network trained on a small subset of circuits and their corresponding ground-truth performances. However, current PQAS algorithms train a single predictor to fit the entire circuit space using limited training samples, which is inefficient and unnecessary, as QAS aims to identify the optimal quantum circuit. In this paper, we propose a progressive PQAS with active learning (PQAS-AL), which gradually trains more precise predictors for subspaces where high-performance circuits reside. PQAS-AL follows an iterative process from coarse to fine. During the early iterations, coarse predictors are trained to identify relatively good subspaces. As the algorithm iterates, an increasing number of high-performance circuits are added to the training set, resulting in the continuous enhancement of the predictor’s ability to recognize high-performance circuits. Moreover, the iterative training of multiple predictors in PQAS-AL also enables the utilization of newly acquired quantum circuits and their corresponding ground-truth performances, resulting in significantly improved sample efficiency. Unlike current QAS algorithms, PQAS-AL naturally integrates the search and performance evaluation modules. Simulation on the Variational Quantum Eigensolver (VQE) for the Heisenberg model demonstrates that PQAS-AL achieves a 2.8
reduction in labeled quantum circuits for the same performance compared to the current PQAS.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2023. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
