基于HSIC Bottleneck的图对比学习方法研究
Research on Graph Contrastive Learning Method Based on HSIC Bottleneck
为了提高自监督对比学习方法的准确性，本文使用位置编码作为图的初始表示，提出了基于HSIC Bottleneck的对比学习损失函数。在节点分类任务上进行实验，本文提出的图对比学习方法的效果要优于现有的无监督模型，和有监督学习模型相比，也具有一定的竞争力。
In order to improve the accuracy of self supervised contrastive learning methods, this paper uses position encoding as the initial representation of the graph and proposes a contrastive learning loss function based on HSIC Bottleneck. Experiments were conducted on node classification tasks, and the proposed graph contrastive learning method outperformed existing unsupervised models. Compared with supervised learning models, it also has certain competitiveness
崔琪
计算技术、计算机技术
计算机软件与理论图对比学习位置编码信息瓶颈
omputer software and theoryGraph contrastive learningLocation codingInformation bottleneck
崔琪.基于HSIC Bottleneck的图对比学习方法研究[EB/OL].(2024-05-16)[2025-08-21].http://www.paper.edu.cn/releasepaper/content/202405-86.点此复制
评论