Link Prediction in Multilayer Networks via Cross-Network Embedding
Authors: Guojing Ren, Xiao Ding, Xiao-Ke Xu, Hai-Feng Zhang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments conducted on real-world datasets demonstrate the superior performance of our proposed method for link prediction in multilayer networks. |
| Researcher Affiliation | Academia | 1Institutes of Physical Science and Information Technology, Anhui University 2Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education, School of Mathematical Science, Anhui University 3 Computational Communication Research Center and School of Journalism and Communication, Beijing Normal University |
| Pseudocode | Yes | Algorithm 1: CGNN |
| Open Source Code | No | The paper does not provide an explicit statement or link for open-source code for the described methodology. |
| Open Datasets | Yes | Datasets We select two real-world datasets: (i) Facebook/Twitter (Du et al. 2022); (ii) Twitter-You Tube (Dickison, Magnani, and Rossi 2016). |
| Dataset Splits | No | The paper states: 'For each dataset, we randomly select 90% edges as the training set, and the rest 10% edges as the test set.' It does not specify a validation set split. |
| Hardware Specification | Yes | Hardware SVD, GAE, GAT, and CGNN were run on a Linux server with an NVIDIA A100-40G GPU. |
| Software Dependencies | No | The paper states: 'These codes are implemented in Python with Py Torch and Py Torch Geometric libraries.' However, it does not provide specific version numbers for these software dependencies. |
| Experiment Setup | Yes | Parameter Setup For SVD, the embedding dimension is 32. For node2vec and n2v-e, p = 1, q = 1, window size is 10, the number of walks per node is 20, walk length is 80, and the embedding dimension is 128. For GAE, the hidden dimension is 32, the embedding dimension is 16, and learning rate is 0.01. For GAT, the first layer consists of 4 attention heads computing 8 features each (for a total of 32 features), the embedding dimension is 16, and learning rate is 0.01. For CGNN, the hidden dimension is 256, the embedding dimension d = 128, α = 0.05, ϵ = 0.7, window size is 10, walks per node is 10, walk length is 20, batch size br = 512, and learning rate is 0.001. |