On the Scalable Learning of Stochastic Blockmodel
Authors: Bo Yang, Xuehua Zhao
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Its effectiveness and efficiency have been tested through rigorous comparisons with the state-of-the-art methods on both synthetic and real-world networks. Next, we design experiments oriented toward evaluating the accuracy, the scalability, and the tradeoff between accuracy and scalability of BLOS, Four state-of-the-art SBM learning methods, VBMOD (Hofman and Wiggins 2008), GSMDL (Yang, Liu, and Liu 2012), SICL (Daudin, Picard, and Robin 2008) and SILvb (Latouche, Birmele, and Ambroise 2012), are selected as comparative methods, whose rationale and time complexity are summarized in Table 2. |
| Researcher Affiliation | Academia | Bo Yang and Xuehua Zhao School of Computer Science and Technology, Jilin University, Changchun, China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, China ybo@jlu.edu.cn |
| Pseudocode | Yes | Table 1: The implementation of block-wise SBM learning |
| Open Source Code | No | The paper does not provide any specific repository link, explicit code release statement, or mention code in supplementary materials for the methodology described. |
| Open Datasets | Yes | Now we test the performance of algorithms with real-world networks. Total 9 real-world networks are selected, which are widely used as benchmarks to validate the performance of block structure detection or scalability. The structural features of these networks are summarized in Table 5. Some have ground truth block structures. Karate Undirected 34 78 0.57 4.59 community Dolphins Undirected 62 159 0.26 5.13 community Foodweb Undirected 75 113 0.33 3.01 hybrid Polbooks Undirected 105 441 0.49 8.40 community Adjnoun Undirected 112 425 0.17 7.59 bipartite Football Undirected 115 613 0.40 10.7 community Email Undirected 1133 5451 0.22 9.62 Polblogs Directed 1222 16714 0.32 27.4 Yeast Undirected 2224 6609 0.13 5.94 |
| Dataset Splits | No | The paper describes generating synthetic networks and using real-world benchmark datasets, but it does not provide specific train/validation/test dataset splits, exact percentages, sample counts, or detailed splitting methodology. |
| Hardware Specification | Yes | All experiments are performed on a conventional personal computer with a 2GH CPU and a 4GB RAM. |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library or solver names with version numbers. |
| Experiment Setup | Yes | For all five algorithms, we set the same model space to search and the same convergence threshold, i.e., Kmin = 1, Kmax = 20 and ε = 10 4. |