Zhang, Yi, Wang, Lulu, Wilson, Richard Charles orcid.org/0000-0001-7265-3033 et al. (1 more author) (Accepted: 2020) An R-convolution Graph Kernel based on Fast Discrete-Time Quantum Walk. IEEE Transactions on Neural Networks and Learning Systems. ISSN 2162-237X (In Press)
Abstract
In this paper, a novel R-convolution kernel, named the fast quantum walk kernel (FQWK), is proposed for unattributed graphs. In FQWK, the similarity of the neighborhood-pair substructure between two nodes is measured via the superposition amplitude of quantum walks between those nodes. The quantum interference in this kind of local substructures provides more information on the substructures so that FQWK can capture finer-grained local structural features of graphs. In addition, to efficiently compute the transition amplitudes of multi-step discrete-time quantum walks, a fast recursive method is designed. Thus compared with all the existing kernels based on the quantum walk, FQWK has the highest computation speed. Extensive experiments demonstrate that FQWK outperforms state-of-the-art graph kernels in terms of classification accuracy for unattributed graphs. Meanwhile, it can be applied to distinguish a larger family of graphs including cospectral graphs, regular graphs, and even strong regular graphs which are not distinguishable by classical walkbased methods.
Metadata
Item Type: | Article |
---|---|
Authors/Creators: |
|
Copyright, Publisher and Additional Information: | © IEEE 2020. This is an author-produced version of the published paper. Uploaded in accordance with the publisher’s self-archiving policy. Further copying may not be permitted; contact the publisher for details. |
Dates: |
|
Institution: | The University of York |
Academic Units: | The University of York > Faculty of Sciences (York) > Computer Science (York) |
Depositing User: | Pure (York) |
Date Deposited: | 30 Sep 2020 10:00 |
Last Modified: | 15 Mar 2025 00:09 |
Status: | In Press |
Refereed: | Yes |
Open Archives Initiative ID (OAI ID): | oai:eprints.whiterose.ac.uk:166053 |