Negative samples selecting strategy for graph contrastive learning

Graph neural networks (GNNs) have emerged as a successful method on graph structured data. Limited by expensive labeled data, contrastive learning has been adopted to the graph domain. In most existing node-level graph contrastive learning methods, when applying contrastive learning to a certain unlabeled node (the center node), its corresponding sim-ilar node (positive sample) is usually generated by data augmentation. Other nodes in the graph are served as the dissimilar nodes (negative samples), which leads to two major problems. First, the computational cost can be prohibitively expensive, especially when the graph is large. Second, utilizing some nodes which share the same label with the center node as the negative samples will damage the learning process. Hence, to address these issues, we explore the feasibility of only sampling a part of nodes for graph contrastive learning process. And unlike the previous self-supervised contrastive methods, we use joint training to exploit supervised signals as much as possible in contrastive learning. Hence, we propose a Negative Samples Selecting Strategy to utilize the classification prediction to guide the selection of the negative samples for sampled nodes. Then, we further incor-porate this strategy for performing contrastive learning on graphs and propose a frame-work named Graph Contrastive Learning with Negative Samples Selecting Strategy (GCNSS). We demonstrate that GCNSS can be trained much faster with much less compu-tation memory than graph contrastive learning baselines, and GCNSS can effectively boost the performance of existing GNN models on semi-supervised node classification tasks across many different datasets. The code is in: https://github.com/MR9812/GCNSS.(c) 2022 Published by Elsevier Inc.

Automated summary

Graph neural networks (GNNs) have been successful in handling graph structured data. Due to limited labeled data, contrastive learning is applied to the graph domain. Existing node-level graph contrastive learning methods face challenges of high computational cost and inaccurate negative sample selection. To address these challenges, we propose a strategy for sampling a subset of nodes and utilize classification prediction to guide the selection of negative samples. Our approach, named GCNSS, achieves faster training and improves the performance of existing GNN models in semi-supervised node classification tasks.