Attribute-Missing Graph Clustering Network

Authors: Wenxuan Tu, Renxiang Guan, Sihang Zhou, Chuan Ma, Xin Peng, Zhiping Cai, Zhe Liu, Jieren Cheng, Xinwang Liu

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on five datasets have verified the superiority of AMGC against competitors.
Researcher Affiliation Academia 1School of Computer, National University of Defense Technology, Changsha, China 2School of Intelligence Science and Technology, National University of Defense Technology, Changsha, China 3Zhejiang Lab, Hangzhou, China 4School of Computer Science and Technology, Hainan University, Haikou, China 5Hainan Blockchain Technology Engineering Research Center, Haikou, China
Pseudocode Yes Algorithm 1: The learning procedure of AMGC
Open Source Code Yes 1https://github.com/Wx Tu/AMGC
Open Datasets Yes To evaluate the proposed AMGC, we choose five public graph datasets, i.e., small-scale Cora and Citeseer, medium-scale Pubmed and Coauthor CS (Co.CS), and large-scale Coauthor Physics (Co.Physics).
Dataset Splits No The paper mentions generating attribute-missing settings by selecting 40% complete and removing 60% attributes. It does not explicitly state the training, validation, and test dataset splits with percentages or sample counts.
Hardware Specification Yes OOM means the out-of-memory failure on 24GB RTX 3090 GPU and 64G RAM.
Software Dependencies No The paper mentions using an 'Adam optimizer' but does not specify version numbers for any software dependencies or libraries.
Experiment Setup Yes According to the results of hyper-parameter sensitivity testing, we set the nearest neighbors K and the edge-masking ratio re as 2 and 0.7, respectively. Moreover, the learning rate, the latent dimension, the dropout rate, and the weight decay are set to 1e-3, 256, 0.4, and 5e-4 as default, respectively.