Power up! Robust graph convolutional network via graph powering. Jin, M., Chang, H., Zhu, W., & Sojoudi, S. In AAAI Conference on Artificial Intelligence (AAAI), 2021. ![Power up! Robust graph convolutional network via graph powering [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Pdf ![Power up! Robust graph convolutional network via graph powering [link]](https://bibbase.org/img/filetypes/link.svg "link")
Arxiv abstract bibtex 8 downloads Graph convolutional networks (GCNs) are powerful tools for graph-structured data. However, they have been recently shown to be vulnerable to topological attacks. To enhance adversarial robustness, we go beyond spectral graph theory to robust graph theory. By challenging the classical graph Laplacian, we propose a new convolution operator that is provably robust in the spectral domain and is incorporated in the GCN architecture to improve expressivity and interpretability. By extending the original graph to a sequence of graphs, we also propose a robust training paradigm that encourages transferability across graphs that span a range of spatial and spectral characteristics. The proposed approaches are demonstrated in extensive experiments to simultaneously improve performance in both benign and adversarial situations.
@inproceedings{2021_3C_power,
title={Power up! Robust graph convolutional network via graph powering},
author={Jin, Ming and Chang, Heng and Zhu, Wenwu and Sojoudi, Somayeh},
booktitle={AAAI Conference on Artificial Intelligence (AAAI)},
url_pdf={Robust_GCN.pdf},
url_arXiv={https://arxiv.org/abs/1905.10029},
keywords={Graph theory, Machine learning},
abstract={Graph convolutional networks (GCNs) are powerful tools for graph-structured data. However, they have been recently shown to be vulnerable to topological attacks. To enhance adversarial robustness, we go beyond spectral graph theory to robust graph theory. By challenging the classical graph Laplacian, we propose a new convolution operator that is provably robust in the spectral domain and is incorporated in the GCN architecture to improve expressivity and interpretability. By extending the original graph to a sequence of graphs, we also propose a robust training paradigm that encourages transferability across graphs that span a range of spatial and spectral characteristics. The proposed approaches are demonstrated in extensive experiments to simultaneously improve performance in both benign and adversarial situations.},
year={2021}
}
Downloads: 8
{"_id":"RPRKG2ohXm3zuYwcw","bibbaseid":"jin-chang-zhu-sojoudi-poweruprobustgraphconvolutionalnetworkviagraphpowering-2021","author_short":["Jin, M.","Chang, H.","Zhu, W.","Sojoudi, S."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"Power up! Robust graph convolutional network via graph powering","author":[{"propositions":[],"lastnames":["Jin"],"firstnames":["Ming"],"suffixes":[]},{"propositions":[],"lastnames":["Chang"],"firstnames":["Heng"],"suffixes":[]},{"propositions":[],"lastnames":["Zhu"],"firstnames":["Wenwu"],"suffixes":[]},{"propositions":[],"lastnames":["Sojoudi"],"firstnames":["Somayeh"],"suffixes":[]}],"booktitle":"AAAI Conference on Artificial Intelligence (AAAI)","url_pdf":"Robust_GCN.pdf","url_arxiv":"https://arxiv.org/abs/1905.10029","keywords":"Graph theory, Machine learning","abstract":"Graph convolutional networks (GCNs) are powerful tools for graph-structured data. However, they have been recently shown to be vulnerable to topological attacks. To enhance adversarial robustness, we go beyond spectral graph theory to robust graph theory. By challenging the classical graph Laplacian, we propose a new convolution operator that is provably robust in the spectral domain and is incorporated in the GCN architecture to improve expressivity and interpretability. By extending the original graph to a sequence of graphs, we also propose a robust training paradigm that encourages transferability across graphs that span a range of spatial and spectral characteristics. The proposed approaches are demonstrated in extensive experiments to simultaneously improve performance in both benign and adversarial situations.","year":"2021","bibtex":"@inproceedings{2021_3C_power,\n title={Power up! Robust graph convolutional network via graph powering},\n author={Jin, Ming and Chang, Heng and Zhu, Wenwu and Sojoudi, Somayeh},\n booktitle={AAAI Conference on Artificial Intelligence (AAAI)},\n url_pdf={Robust_GCN.pdf},\n url_arXiv={https://arxiv.org/abs/1905.10029},\n keywords={Graph theory, Machine learning},\n abstract={Graph convolutional networks (GCNs) are powerful tools for graph-structured data. However, they have been recently shown to be vulnerable to topological attacks. To enhance adversarial robustness, we go beyond spectral graph theory to robust graph theory. By challenging the classical graph Laplacian, we propose a new convolution operator that is provably robust in the spectral domain and is incorporated in the GCN architecture to improve expressivity and interpretability. By extending the original graph to a sequence of graphs, we also propose a robust training paradigm that encourages transferability across graphs that span a range of spatial and spectral characteristics. The proposed approaches are demonstrated in extensive experiments to simultaneously improve performance in both benign and adversarial situations.},\n year={2021}\n}\n","author_short":["Jin, M.","Chang, H.","Zhu, W.","Sojoudi, S."],"key":"2021_3C_power","id":"2021_3C_power","bibbaseid":"jin-chang-zhu-sojoudi-poweruprobustgraphconvolutionalnetworkviagraphpowering-2021","role":"author","urls":{" pdf":"http://www.jinming.tech/papers/Robust_GCN.pdf"," arxiv":"https://arxiv.org/abs/1905.10029"},"keyword":["Graph theory","Machine learning"],"metadata":{"authorlinks":{}},"downloads":8},"bibtype":"inproceedings","biburl":"http://www.jinming.tech/papers/myref.bib","dataSources":["sTzDHHaipTZWjp8oe","Y64tp2HnDCfXgLdc5"],"keywords":["graph theory","machine learning"],"search_terms":["power","robust","graph","convolutional","network","via","graph","powering","jin","chang","zhu","sojoudi"],"title":"Power up! Robust graph convolutional network via graph powering","year":2021,"downloads":8}