amyisnotbusy shen | Hefei University of technology (original) (raw)
Related Authors
Rajshahi University of Engineering and Technology
Uploads
Papers by amyisnotbusy shen
Geometric deep learning, a novel class of machine learning algorithms extending classical deep le... more Geometric deep learning, a novel class of machine learning algorithms extending classical deep learning architectures to non-Euclidean structured data such as manifolds and graphs, has recently been applied to a broad spectrum of problems ranging from computer graphics and chemistry to high energy physics and social media. Despite the plenitude of different graph neural network architectures that have been proposed, until recently relatively little effort has been dedicated to developing methods that scale to very large graphs -which precludes their industrial applications e.g. in social networks. In this paper, we propose SIGN, a scalable graph neural network analogous to the popular inception module used in classical convolutional architectures. We show that our architecture is able to effectively deal with large-scale graphs via pre-computed multi-scale neighborhood features. Extensive experimental evaluation on various open benchmarks shows the competitive performance of our approach compared to a variety of popular architectures, while requiring a fraction of training and inference time. * Equal contribution Preprint. Under review.
Geometric deep learning, a novel class of machine learning algorithms extending classical deep le... more Geometric deep learning, a novel class of machine learning algorithms extending classical deep learning architectures to non-Euclidean structured data such as manifolds and graphs, has recently been applied to a broad spectrum of problems ranging from computer graphics and chemistry to high energy physics and social media. Despite the plenitude of different graph neural network architectures that have been proposed, until recently relatively little effort has been dedicated to developing methods that scale to very large graphs -which precludes their industrial applications e.g. in social networks. In this paper, we propose SIGN, a scalable graph neural network analogous to the popular inception module used in classical convolutional architectures. We show that our architecture is able to effectively deal with large-scale graphs via pre-computed multi-scale neighborhood features. Extensive experimental evaluation on various open benchmarks shows the competitive performance of our approach compared to a variety of popular architectures, while requiring a fraction of training and inference time. * Equal contribution Preprint. Under review.