SCE: Scalable Network Embedding from Sparsest Cut Article Swipe
YOU?
·
· 2020
· Open Access
·
· DOI: https://doi.org/10.1145/3394486.3403068
Large-scale network embedding is to learn a latent representation for each\nnode in an unsupervised manner, which captures inherent properties and\nstructural information of the underlying graph. In this field, many popular\napproaches are influenced by the skip-gram model from natural language\nprocessing. Most of them use a contrastive objective to train an encoder which\nforces the embeddings of similar pairs to be close and embeddings of negative\nsamples to be far. A key of success to such contrastive learning methods is how\nto draw positive and negative samples. While negative samples that are\ngenerated by straightforward random sampling are often satisfying, methods for\ndrawing positive examples remains a hot topic.\n In this paper, we propose SCE for unsupervised network embedding only using\nnegative samples for training. Our method is based on a new contrastive\nobjective inspired by the well-known sparsest cut problem. To solve the\nunderlying optimization problem, we introduce a Laplacian smoothing trick,\nwhich uses graph convolutional operators as low-pass filters for smoothing node\nrepresentations. The resulting model consists of a GCN-type structure as the\nencoder and a simple loss function. Notably, our model does not use positive\nsamples but only negative samples for training, which not only makes the\nimplementation and tuning much easier, but also reduces the training time\nsignificantly.\n Finally, extensive experimental studies on real world data sets are\nconducted. The results clearly demonstrate the advantages of our new model in\nboth accuracy and scalability compared to strong baselines such as GraphSAGE,\nG2G and DGI.\n
Related Topics To Compare & Contrast
- Type
- preprint
- Language
- en
- Landing Page
- https://doi.org/10.1145/3394486.3403068
- OA Status
- green
- Cited By
- 7
- References
- 50
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W3038832843