Dynamic Network Embedding by Modeling Triadic Closure Process
Authors: Lekui Zhou, Yang Yang, Xiang Ren, Fei Wu, Yueting Zhuang
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on three real-world networks demonstrate that, compared with several state-of-the-art techniques, Dynamic Triad achieves substantial gains in several application scenarios. |
| Researcher Affiliation | Academia | Lekui Zhou,1 Yang Yang,1 Xiang Ren,2 Fei Wu,1 Yueting Zhuang1 1 Department of Computer Science and Technology, Zhejiang University 2 Department of Computer Science, University of Southern California {luckiezhou, yangya, wufei, yzhuang}@zju.edu.cn, xiangren@usc.edu |
| Pseudocode | Yes | Algorithm 1: Training process of Dynamic Triad |
| Open Source Code | Yes | An implementation of the model is publicly available 1. 1https://github.com/luckiezhou/Dynamic Triad |
| Open Datasets | Yes | Academic. We derive a co-authorship network from the academic network of AMiner4 (Tang et al. 2008). (Footnote 4: http://www.aminer.org, an academic search engine) |
| Dataset Splits | Yes | Using a Logisitic Regression model as classifier, we repeat 5-fold cross validation on the gathered sample set for 10 times, and compare the average performance. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware used to run the experiments (e.g., CPU/GPU models, memory specifications). |
| Software Dependencies | No | The paper mentions using a 'Logisitic Regression model' and 'SGD framework with Adagrad method' but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | We have conducted our experiments on dimensions 16, 32, 48 and 64, and we have chosen d = 48 for presentation due to space limitation. For each graph embedding algorithm mentioned in this section, we perform a grid search on the values of hyper parameters, and we choose a specific combination of them for each task on each data set, which results in the best performance regarding to the F1 score metric. We tested all combinations of parameters β0 and β1 given β0 {0.01, 0.1, 1, 10} and β1 {0.01, 0.1, 1, 10}. |