Convolutional Kernel Networks for Graph-Structured Data

Dexiong Chen, Laurent Jacob, Julien Mairal
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:1576-1586, 2020.

Abstract

We introduce a family of multilayer graph kernels and establish new links between graph convolutional neural networks and kernel methods. Our approach generalizes convolutional kernel networks to graph-structured data, by representing graphs as a sequence of kernel feature maps, where each node carries information about local graph substructures. On the one hand, the kernel point of view offers an unsupervised, expressive, and easy-to-regularize data representation, which is useful when limited samples are available. On the other hand, our model can also be trained end-to-end on large-scale data, leading to new types of graph convolutional neural networks. We show that our method achieves competitive performance on several graph classification benchmarks, while offering simple model interpretation. Our code is freely available at https://github.com/claying/GCKN.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-chen20h, title = {Convolutional Kernel Networks for Graph-Structured Data}, author = {Chen, Dexiong and Jacob, Laurent and Mairal, Julien}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {1576--1586}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/chen20h/chen20h.pdf}, url = {https://proceedings.mlr.press/v119/chen20h.html}, abstract = {We introduce a family of multilayer graph kernels and establish new links between graph convolutional neural networks and kernel methods. Our approach generalizes convolutional kernel networks to graph-structured data, by representing graphs as a sequence of kernel feature maps, where each node carries information about local graph substructures. On the one hand, the kernel point of view offers an unsupervised, expressive, and easy-to-regularize data representation, which is useful when limited samples are available. On the other hand, our model can also be trained end-to-end on large-scale data, leading to new types of graph convolutional neural networks. We show that our method achieves competitive performance on several graph classification benchmarks, while offering simple model interpretation. Our code is freely available at https://github.com/claying/GCKN.} }
Endnote
%0 Conference Paper %T Convolutional Kernel Networks for Graph-Structured Data %A Dexiong Chen %A Laurent Jacob %A Julien Mairal %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-chen20h %I PMLR %P 1576--1586 %U https://proceedings.mlr.press/v119/chen20h.html %V 119 %X We introduce a family of multilayer graph kernels and establish new links between graph convolutional neural networks and kernel methods. Our approach generalizes convolutional kernel networks to graph-structured data, by representing graphs as a sequence of kernel feature maps, where each node carries information about local graph substructures. On the one hand, the kernel point of view offers an unsupervised, expressive, and easy-to-regularize data representation, which is useful when limited samples are available. On the other hand, our model can also be trained end-to-end on large-scale data, leading to new types of graph convolutional neural networks. We show that our method achieves competitive performance on several graph classification benchmarks, while offering simple model interpretation. Our code is freely available at https://github.com/claying/GCKN.
APA
Chen, D., Jacob, L. & Mairal, J.. (2020). Convolutional Kernel Networks for Graph-Structured Data. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:1576-1586 Available from https://proceedings.mlr.press/v119/chen20h.html.

Related Material