Non-IID Transfer Learning on Graphs
Authors: Jun Wu, Jingrui He, Elizabeth Ainsworth
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results verify the effectiveness and efficiency of our GRADE framework on both cross-network node classification and cross-domain recommendation tasks. ... Experiment Experimental Setup Data Sets ... Baselines ... Model Configuration ... Cross-Network Node Classification Table 1 and Table 2 provide the cross-network node classification results of GRADE-N. ... Cross-Domain Recommendation Results on overlapping users Table 4 provides the cross-domain recommendation results on the Amazon data set. ... Flexibility Figure 3 shows the performance of GRADE-N with different base discrepancies and base graph neural network architectures on social networks. ... Hyper-parameter Sensitivity We investigate the impact of λ on GRADE-N. ... Computational Efficiency We investigate the computational efficiency of GRADE framework. |
| Researcher Affiliation | Academia | Jun Wu1, Jingrui He1, Elizabeth Ainsworth1,2 1University of Illinois Urbana-Champaign 2USDA ARS Global Change and Photosynthesis Research Unit {junwu3, jingrui, ainswort}@illinois.edu |
| Pseudocode | No | The paper describes its proposed algorithms and framework using mathematical formulations and textual explanations, but it does not provide any explicitly labeled 'Pseudocode' or 'Algorithm' blocks with structured steps. |
| Open Source Code | Yes | 3https://github.com/jwu4sml/GRADE |
| Open Datasets | Yes | Data Sets For cross-network node classification, we use the following data sets: Airport networks (Ribeiro, Saverese, and Figueiredo 2017) (Brazil, USA and Europe); Citation network (Wu et al. 2020; Tang et al. 2008) (ACMv9 (A) and DBLPv8 (D)); Social network (Shen et al. 2020; Li et al. 2015) (Blog1 (B1) and Blog2 (B2)); and Agriculture data (Wang et al. 2021) (Maize (M) and Maize UNL (MU)). For cross-domain recommendation, we evaluate the models on the Amazon data set (He and Mc Auley 2016). |
| Dataset Splits | No | The paper uses standard datasets and mentions evaluation on target graphs but does not provide explicit training, validation, and test split percentages or sample counts, nor does it refer to predefined benchmark splits that include validation subsets. |
| Hardware Specification | No | The paper discusses computational efficiency and running time but does not provide specific hardware details such as GPU or CPU models, memory, or cloud instance types used for the experiments. |
| Software Dependencies | No | The paper mentions using Graph Convolutional Network (GCN) and multi-layer perceptron (MLP) models, but it does not provide specific version numbers for any software dependencies or libraries used for implementation. |
| Experiment Setup | Yes | Model Configuration We adopt two hidden layers in the base GCN model when implementing the GRADE algorithms. We set λ = 0.02 for cross-network node classification and λ = 0.1 for the cross-domain recommendation. |