Lagrangian Constrained Community Detection

Authors: Mohadeseh Ganji, James Bailey, Peter Stuckey

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on real and synthetic data sets show our proposed Lag CCD algorithm outperforms existing algorithms in terms of solution quality, ability to satisfy the constraints and noise resistance.
Researcher Affiliation Academia Mohadeseh Ganj, James Bailey, Peter J. Stuckey School of Computing and Information Systems, University of Melbourne, Australia sghasempour@student.unimelb.edu.au, {baileyj,pstuckey}@unimelb.edu.au
Pseudocode Yes The pseudo-code of the Lag CCD algorithm is shown in Figure 1.
Open Source Code No The paper does not provide any links or explicit statements about the availability of its source code.
Open Datasets Yes The information about the real data sets including their number of vertices (n) and edges (|E|) and number of ground truth communities (k) are shown in Table 1. ... We use LFR data sets proposed in (Lancichinetti, Fortunato, and Radicchi 2008).
Dataset Splits No The paper does not explicitly describe standard training, validation, and test dataset splits. It discusses generating constraints for datasets, but not splitting the data itself for model training/validation phases.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper mentions using 'Gen Louvain (Inderjit S. Jutla and Mucha 2011)' and modifying 'the implementation of (Kuang, Ding, and Park 2012)' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes Our proposed algorithm is denoted by Lag CCD. The parameter α is 1.2 and maximum iteration is set to 30. Violation penalty in P ML and P CL are set to 1. We used Gen Louvain (Inderjit S. Jutla and Mucha 2011) as the optimization step in GD where we used W c as the similarity matrix.