Be Confident! Towards Trustworthy Graph Neural Networks via Confidence Calibration

Authors: Xiao Wang, Hongrui Liu, Chuan Shi, Cheng Yang

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the effectiveness of our proposed model in terms of both calibration and accuracy.
Researcher Affiliation Academia Xiao Wang, Hongrui Liu, Chuan Shi , Cheng Yang School of Computer Science (National Pilot Software Engineering School) Beijing University of Posts and Telecommunications Beijing, China {xiaowang, liuhongrui, shichuan, yangcheng}@bupt.edu.cn
Pseudocode No The paper includes figures illustrating frameworks but does not contain any explicit pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes We choose the commonly used citation networks Cora [29], Citeseer [29], Pubmed [29] and Cora Full [3] for evaluation, and more detailed descriptions are in Appendix B.
Dataset Splits Yes Given an unlabeled dataset DU and a labeled dataset DL which has been divided into three parts Dtrain, Dval and Dtest, we firstly train a classification GCN using Dtrain to get the logit of each node.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory, or cloud instance types).
Software Dependencies No The paper mentions using GCN and GAT, and implementing post-hoc calibration techniques, but does not provide specific version numbers for any software dependencies (e.g., Python, PyTorch, TensorFlow).
Experiment Setup Yes For our Ca GCN, we train a two-layer GCN with the hidden layer dimension to be 16. We set λ = 0.5 for all datasets, weight decay to be 5e-3 for Cora, Citeseer, Pubmed and 0.03 for Cora Full. Other parameters of Ca GCN follows [16]. We set the learning rate lr = 0.001 for Ca GCN-st and train our Ca GCN-st 200 epochs for Cora, 150 epochs for Citeseer, 100 epochs for Pubmed and 500 epochs for Cora Full. We set the threshold τ {0.8,0.85,0.9,0.95,0.99} and the maximum number of stage s = 10.