GraFN: Semi-Supervised Node Classification on Graph with Few Labels via Non-Parametric Distribution Assignment

被引:10
作者
Lee, Junseok [1 ]
Oh, Yunhak [1 ]
In, Yeonjun [1 ]
Lee, Namkyeong [1 ]
Hyun, Dongmin [2 ]
Park, Chanyoung [3 ]
机构
[1] KAIST ISysE, Daejeon, South Korea
[2] POSTECH PIAI, Pohang, South Korea
[3] KAIST ISysE & AI, Daejeon, South Korea
来源
PROCEEDINGS OF THE 45TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '22) | 2022年
关键词
Semi-Supervised; Graph Neural Networks; Node Classification; Graph Representation Learning; Few Label;
D O I
10.1145/3477495.3531838
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Despite the success of Graph Neural Networks (GNNs) on various applications, GNNs encounter significant performance degradation when the amount of supervision signals, i.e., number of labeled nodes, is limited, which is expected as GNNs are trained solely based on the supervision obtained from the labeled nodes. On the other hand, recent self-supervised learning paradigm aims to train GNNs by solving pretext tasks that do not require any labeled nodes, and it has shown to even outperform GNNs trained with few labeled nodes. However, a major drawback of self-supervised methods is that they fall short of learning class discriminative node representations since no labeled information is utilized during training. To this end, we propose a novel semi-supervised method for graphs, GraFN(1) that leverages few labeled nodes to ensure nodes that belong to the same class to be grouped together, thereby achieving the best of both worlds of semi-supervised and self-supervised methods. Specifically, GraFN randomly samples support nodes from labeled nodes and anchor nodes from the entire graph. Then, it minimizes the difference between two predicted class distributions that are non-parametrically assigned by anchor-supports similarity from two differently augmented graphs. We experimentally show that GraFN surpasses both the semi-supervised and self-supervised methods in terms of node classification on real-world graphs.
引用
收藏
页码:2243 / 2248
页数:6
相关论文
共 32 条
[1]  
Arazo E., 2020, IEEE IJCNN
[2]   Semi-Supervised Learning of Visual Features by Non-Parametrically Predicting View Assignments with Support Samples [J].
Assran, Mahmoud ;
Caron, Mathilde ;
Misra, Ishan ;
Bojanowski, Piotr ;
Joulin, Armand ;
Ballas, Nicolas ;
Rabbat, Michael .
2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, :8423-8432
[3]  
Feng WZ, 2020, ADV NEUR IN, V33
[4]   LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation [J].
He, Xiangnan ;
Deng, Kuan ;
Wang, Xiang ;
Li, Yan ;
Zhang, Yongdong ;
Wang, Meng .
PROCEEDINGS OF THE 43RD INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '20), 2020, :639-648
[5]   Residual-Duet Network with Tree Dependency Representation for Chinese Question-Answering Sentiment Analysis [J].
Hu, Guangyi ;
Shi, Chongyang ;
Hao, Shufeng ;
Bai, Yu .
PROCEEDINGS OF THE 43RD INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '20), 2020, :1725-1728
[6]   Multi-behavior Recommendation with Graph Convolutional Networks [J].
Jin, Bowen ;
Gao, Chen ;
He, Xiangnan ;
Jin, Depeng ;
Li, Yong .
PROCEEDINGS OF THE 43RD INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '20), 2020, :659-668
[7]  
Kipf T. N., 2017, ICLR, P1, DOI https://doi.org/10.48550/arXiv.1609.02907
[8]  
Klicpera J., 2018, ICLR
[9]  
Li Q, 2019, INT GEOSCI REMOTE SE, P9585, DOI 10.1109/IGARSS.2019.8897919
[10]  
Li QM, 2018, AAAI CONF ARTIF INTE, P3538