Scalable and Effective Conductance-Based Graph Clustering

Authors: Longlong Lin, Ronghua Li, Tao Jia

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results on real-life and synthetic datasets show that our algorithms can achieve 5 42 times speedup with a high clustering accuracy, while using 1.4 7.8 times less memory than the baseline algorithms.
Researcher Affiliation Academia Longlong Lin1, Rong-Hua Li 2, Tao Jia1 1College of Computer and Information Science, Southwest University, Chongqing 400715, China 2Beijing Institute of Technology, China
Pseudocode Yes Algorithm 1: PCon core; Algorithm 2: PCon de
Open Source Code No The paper does not provide any links to the source code for the described methodology or state that the code is being released.
Open Datasets Yes We evaluate our proposed solutions on six real-life publiclyavailable datasets2 (Table 2), which are widely used benchmarks for conductance-based graph clustering (Shun et al. 2016; Yang et al. 2019). The maximum connected components of these datasets are used in the experiments. We also use five synthetic graphs LFR (Lancichinetti, Fortunato, and Kert esz 2009), WS (Watts and Strogatz 1998), PLC (Holme and Kim 2002), ER (Erdos, R enyi et al. 1960), and BA (Barab asi and Albert 1999). All datasets can be downloaded from http://snap.stanford.edu/
Dataset Splits No The paper uses various datasets for experiments but does not explicitly describe any train/validation/test splits, nor does it refer to standard predefined splits for the specific datasets used.
Hardware Specification No The paper does not provide any specific hardware details such as CPU/GPU models, memory, or cloud instance types used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers for its implementation. It mentions parameters for baseline methods, but not for the proposed algorithms' software environment.
Experiment Setup No The paper provides specific parameter settings (e.g., α = 0.01 and ϵ = 1 m for NIBBLE PPR; t = 10 and ϵ = 1 m for HK Relax) for baseline algorithms, but does not specify experimental setup details such as hyperparameters or training configurations for the proposed PCon core and PCon de algorithms.