Rayleigh Quotient Graph Neural Networks for Graph-level Anomaly Detection

Authors: Xiangyu Dong, Xingyi Zhang, Sibo Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on 10 real-world datasets show that RQGNN outperforms the best rival by 6.74% in Macro-F1 score and 1.44% in AUC, demonstrating the effectiveness of our framework.
Researcher Affiliation Academia Xiangyu Dong, Xingyi Zhang, Sibo Wang Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong {xydong, xyzhang, swang}@se.cuhk.edu.hk
Pseudocode Yes Algorithm 1: RQL Algorithm 2: CWGNN with RQ-pooling Algorithm 3: RQGNN
Open Source Code Yes Our code is available at https://github.com/xydong127/RQGNN.
Open Datasets Yes Datasets. We use 10 real-world datasets to investigate the performance of RQGNN, including MCF7, MOLT-4, PC-3, SW-620, NCI-H23, OVCAR-8, P388, SF-295, SN12C, and UACC257. These datasets are obtained from the TUDataset (Morris et al., 2020), consisting of various chemical compounds and their reactions to different cancer cells.
Dataset Splits Yes Experimental Settings. We randomly divide each dataset into training/validation/test sets with 70%/15%/15%, respectively. During the sampling process, we ensure that each set maintains a consistent ratio between normal and anomalous graphs.
Hardware Specification No No specific hardware details (e.g., GPU models, CPU types, memory) used for running the experiments are mentioned in the paper.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions) are mentioned in the paper.
Experiment Setup Yes Experimental Settings. We set the learning rate as 0.005, the batch size as 512, the hidden dimension d = 64, the width of CWGNN-RQ q = 4, the depth of CWGNN-RQ K = 6, the dropout rate as 0.4, the hyperparameters of the loss function β = 0.999, γ = 1.5, and we use batch normalization for the final graph embeddings.