Rethinking Propagation for Unsupervised Graph Domain Adaptation

Authors: Meihan Liu, Zeyu Fang, Zhen Zhang, Ming Gu, Sheng Zhou, Xin Wang, Jiajun Bu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive experiments on real-world datasets, we demonstrate the effectiveness of our proposed A2GNN framework. Experiments and Analyses Datasets We conduct comprehensive experiments on three public citation networks and two social networks across a range of settings.
Researcher Affiliation Academia Meihan Liu1, Zeyu Fang1, Zhen Zhang3, Ming Gu1, Sheng Zhou1,2*, Xin Wang4, Jiajun Bu1 1Zhejiang Provincial Key Laboratory of Service Robot, College of Computer Science, Zhejiang University 2School of Software Technology, Zhejiang University 3Department of Computer Science, National University of Singapore 4Department of Computer Science and Technology, Tsinghua University
Pseudocode No The paper describes the proposed framework and its components using text and mathematical equations, but it does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes 1https://github.com/Meihan-Liu/24AAAI-A2GNN
Open Datasets Yes We conduct comprehensive experiments on three public citation networks and two social networks across a range of settings. These datasets are acquired from diverse sources and time periods with explicit covariate shifts. The citation networks2 consist of three datasets: ACMv9, Citationv1, and DBLPv7... 2https://github.com/yuntaodu/ASN/tree/main/data and As for social networks, we choose Twitch gamer networks3, which are collected from different regions... 3http://snap.stanford.edu/data/twitch-social-networks.html
Dataset Splits Yes Among them, 80% of the labelled source nodes are utilized to provide supervision signals. The remaining labelled source nodes compose the validation set, and the evaluation is carried out on the target nodes.
Hardware Specification No The paper mentions software implementation details such as PyTorch, but it does not provide any specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No Our proposed A2GNN is implemented with PyTorch (Paszke et al. 2019), where the learning rate and weight decay are searched in the range of {1e 1, 1e 2, 1e 3, 1e 4, 5e 4}. The paper mentions PyTorch but does not specify its version number or any other software dependencies with their versions.
Experiment Setup Yes The node representation dimension is uniformly set as 128 for all the methods. Our proposed A2GNN is implemented with PyTorch (Paszke et al. 2019), where the learning rate and weight decay are searched in the range of {1e 1, 1e 2, 1e 3, 1e 4, 5e 4}. The experiments are repeated five times, and we report the mean performance in terms of Micro-F1 and Macro-F1 scores.