DropEdge: Towards Deep Graph Convolutional Networks on Node Classification
Authors: Yu Rong, Wenbing Huang, Tingyang Xu, Junzhou Huang
ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on several benchmarks verify that Drop Edge consistently improves the performance on a variety of both shallow and deep GCNs. |
| Researcher Affiliation | Collaboration | Yu Rong1, Wenbing Huang2 , Tingyang Xu1, Junzhou Huang1 1 Tencent AI Lab 2 Beijing National Research Center for Information Science and Technology (BNRist), State Key Lab on Intelligent Technology and Systems, Department of Computer Science and Technology, Tsinghua University |
| Pseudocode | No | The paper describes the methodology using text and mathematical equations, but it does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Codes are released on https://github.com/Drop Edge/Drop Edge. |
| Open Datasets | Yes | We focus on four benchmark datasets varying in graph size and feature type: (1) classifying the research topic of papers in three citation datasets: Cora, Citeseer and Pubmed (Sen et al., 2008); (2) predicting which community different posts belong to in the Reddit social network (Hamilton et al., 2017). The statics of all datasets are listed in the supplemental materials. |
| Dataset Splits | Yes | The statics of all datasets are summarized in Table 3. ... Cora 2,708 Nodes 5,429 Edges 7 Classes 1,433 Features 1,208/500/1,000 Transductive |
| Hardware Specification | Yes | All experiments are conducted on a NVIDIA Tesla P40 GPU with 24GB memory. |
| Software Dependencies | No | All backbones are implemented in Pytorch (Paszke et al., 2017). For Graph SAGE, we utilize the Pytorch version implemented by DGL(Wang et al., 2019). No specific version numbers for Pytorch or DGL are provided. |
| Experiment Setup | Yes | We adopt the Adam optimizer for model training. ... We fix the number of training epoch to 400 for all datasets. ... Given a model with n {2, 4, 8, 16, 32, 64} layers, the hidden dimension is 128 and we conduct a random search strategy to optimize the other hyper-parameter for each backbone in 5.1. The decryptions of hyper-parameters are summarized in Table 4. ... Table 6 summaries the hyper-parameters of each backbone with the best accuracy on different datasets and their best accuracy are reported in Table 2. |