On the stability of graph convolutional neural networks under edge rewiring
Graph neural networks are experiencing a surge of popularity within the machine learning community due to their ability to adapt to non-Euclidean domains and instil inductive biases. Despite this, their stability, i.e., their robustness to small perturbations in the input, is not yet well understood...
| Main Authors: | , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
IEEE
2021
|
| _version_ | 1826310999513235456 |
|---|---|
| author | Kenlay, H Thanou, D Dong, X |
| author_facet | Kenlay, H Thanou, D Dong, X |
| author_sort | Kenlay, H |
| collection | OXFORD |
| description | Graph neural networks are experiencing a surge of popularity within the machine learning community due to their ability to adapt to non-Euclidean domains and instil inductive biases. Despite this, their stability, i.e., their robustness to small perturbations in the input, is not yet well understood. Although there exists some results showing the stability of graph neural networks, most take the form of an upper bound on the magnitude of change due to a perturbation in the graph topology. However, the change in the graph topology captured in existing bounds tend not to be expressed in terms of structural properties, limiting our understanding of the model robustness properties. In this work, we develop an interpretable upper bound elucidating that graph neural networks are stable to rewiring between high degree nodes. This bound and further research in bounds of similar type provide further understanding of the stability properties of graph neural networks. |
| first_indexed | 2024-03-07T08:01:56Z |
| format | Conference item |
| id | oxford-uuid:5440118a-180e-4d97-bbfa-45435eeb3997 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T08:01:56Z |
| publishDate | 2021 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:5440118a-180e-4d97-bbfa-45435eeb39972023-10-11T09:21:53ZOn the stability of graph convolutional neural networks under edge rewiringConference itemhttp://purl.org/coar/resource_type/c_5794uuid:5440118a-180e-4d97-bbfa-45435eeb3997EnglishSymplectic ElementsIEEE2021Kenlay, HThanou, DDong, XGraph neural networks are experiencing a surge of popularity within the machine learning community due to their ability to adapt to non-Euclidean domains and instil inductive biases. Despite this, their stability, i.e., their robustness to small perturbations in the input, is not yet well understood. Although there exists some results showing the stability of graph neural networks, most take the form of an upper bound on the magnitude of change due to a perturbation in the graph topology. However, the change in the graph topology captured in existing bounds tend not to be expressed in terms of structural properties, limiting our understanding of the model robustness properties. In this work, we develop an interpretable upper bound elucidating that graph neural networks are stable to rewiring between high degree nodes. This bound and further research in bounds of similar type provide further understanding of the stability properties of graph neural networks. |
| spellingShingle | Kenlay, H Thanou, D Dong, X On the stability of graph convolutional neural networks under edge rewiring |
| title | On the stability of graph convolutional neural networks under edge rewiring |
| title_full | On the stability of graph convolutional neural networks under edge rewiring |
| title_fullStr | On the stability of graph convolutional neural networks under edge rewiring |
| title_full_unstemmed | On the stability of graph convolutional neural networks under edge rewiring |
| title_short | On the stability of graph convolutional neural networks under edge rewiring |
| title_sort | on the stability of graph convolutional neural networks under edge rewiring |
| work_keys_str_mv | AT kenlayh onthestabilityofgraphconvolutionalneuralnetworksunderedgerewiring AT thanoud onthestabilityofgraphconvolutionalneuralnetworksunderedgerewiring AT dongx onthestabilityofgraphconvolutionalneuralnetworksunderedgerewiring |