Quantum machine learning of graph-structured data

authored by: Kerstin Beer, Megha Khosla, Julius Köhler, Tobias J. Osborne, Tianqi Zhao
Abstract: Graph structures are ubiquitous throughout the natural sciences. Here we develop an approach that exploits the quantum source's graph structure to improve learning via an arbitrary quantum neural network (QNN) ansatz. In particular, we devise and optimize a self-supervised objective to capture the information-theoretic closeness of the quantum states in the training of a QNN. Numerical simulations show that our approach improves the learning efficiency and the generalization behavior of the base QNN. On a practical note, scalable quantum implementations of the learning procedure described in this paper are likely feasible on the next generation of quantum computing devices.
Organisation(s): Institute of Theoretical Physics
CRC 1227 Designed Quantum States of Matter (DQ-mat)
External Organisation(s): Macquarie University
Delft University of Technology
Type: Article
Journal: Physical Review A
Volume: 108
ISSN: 2469-9926
Publication date: 10.07.2023
Publication status: Published
Peer reviewed: Yes
ASJC Scopus subject areas: Atomic and Molecular Physics, and Optics
Electronic version(s): https://doi.org/10.48550/arXiv.2103.10837 (Access: Open)
https://doi.org/10.1103/PhysRevA.108.012410 (Access: Closed)

BibTeX

@article{f3a8f50e045549f5a2827ad497c144c4,
title = "Quantum machine learning of graph-structured data",
abstract = "Graph structures are ubiquitous throughout the natural sciences. Here we develop an approach that exploits the quantum source's graph structure to improve learning via an arbitrary quantum neural network (QNN) ansatz. In particular, we devise and optimize a self-supervised objective to capture the information-theoretic closeness of the quantum states in the training of a QNN. Numerical simulations show that our approach improves the learning efficiency and the generalization behavior of the base QNN. On a practical note, scalable quantum implementations of the learning procedure described in this paper are likely feasible on the next generation of quantum computing devices.",
author = "Kerstin Beer and Megha Khosla and Julius K{\"o}hler and Osborne, {Tobias J.} and Tianqi Zhao",
note = "Funding Information: Helpful correspondence and discussions with D. Bondarenko, T. Farrelly, P. Feldmann, A. Hahn, G. M{\"u}ller, J. Hendrik Pfau, R. Salzmann, D. Scheiermann, V. Schmiesing, M. Schwiering, C. Struckmann, and R. Wolf are gratefully acknowledged. This work was supported in part by the Quantum Valley Lower Saxony, the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through SFB 1227 (DQ-mat), the RTG 1991, and DFG under Germany's Excellence Strategy EXC-2123 QuantumFrontiers Grant No. 390837967. ",
year = "2023",
month = jul,
day = "10",
doi = "10.48550/arXiv.2103.10837",
language = "English",
volume = "108",
journal = "Physical Review A",
issn = "2469-9926",
publisher = "American Physical Society",
number = "1",
}