Embedding Compression with Hashing for Efficient Representation Learning in Graph

Chin-Chia Michael Yeh, Mengting Gu, Yan Zheng, Huiyuan Chen, Javid Ebrahimi, Zhongfang Zhuang, Junpeng Wang, Liang Wang, Wei Zhang

Keywords: embedding compression, hashing, graph

Abstract: Graph neural networks (GNNs) are deep learning models designed specifically for graph data, and they typically rely on node features as the input node representation to the first layer. When applying such type of networks on graph without node feature, one can extract simple graph-based node features (e.g., number of degrees) or learn the input node representation (i.e., embeddings) when training the network. While the latter approach, which trains node embeddings, more likely leads to better performance, the number of parameters associated with the embeddings grows linearly with the number of nodes. It is therefore impractical to train the input node embeddings together with GNNs within graphics processing unit (GPU) memory in an end-to-end fashion when dealing with industrial scale graph data. Inspired by the embedding compression methods developed for natural language processing (NLP) models, we develop a node embedding compression method where each node is compactly represented with a bit vector instead of a float-point vector. The parameters utilized in the compression method can be trained together with GNNs. We show that the proposed node embedding compression method achieves superior performance compared to the alternatives.

One-sentence Summary: A embedding compression method is developed to represent each node in a graph compactly for graph neural networks like GraphSage.