Subgraph Federated Learning with Missing Neighbor Generation

Authors: Ke ZHANG, Carl Yang, Xiaoxiao Li, Lichao Sun, Siu Ming Yiu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results on four real-world graph datasets with synthesized subgraph federated learning settings demonstrate the effectiveness and efficiency of our proposed techniques.
Researcher Affiliation Academia Ke Zhang1,4, Carl Yang1 , Xiaoxiao Li2, Lichao Sun3, Siu Ming Yiu4 1Emory University, 2University of British Columbia, 3Lehigh University, 4University of Hong Kong
Pseudocode Yes Appendix A shows the pseudo code of Fed Sage+.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We synthesize the distributed subgraph system with four widely used real-world graph datasets, i.e., Cora [25], Citeseer [25], Pub Med [22], and MSAcademic [26].
Dataset Splits Yes The training-validation-testing ratio is 60%-20%-20% due to limited sizes of local subgraphs.
Hardware Specification Yes We implement Fed Sage and Fed Sage+ in Python and execute all experiments on a server with 8 NVIDIA Ge Force GTX 1080 Ti GPUs.
Software Dependencies No The paper mentions implementing in 'Python' but does not specify version numbers for Python or any other libraries/software dependencies like PyTorch, TensorFlow, etc., that would be needed for replication.
Experiment Setup Yes The number of nodes sampled in each layer of Graph Sage is 5. We use batch size 64 and set training epochs to 50. The training-validation-testing ratio is 60%-20%-20% due to limited sizes of local subgraphs. ... All λs are simply set to 1. Optimization is done with Adam with a learning rate of 0.001.