Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Lower Ricci Curvature for Efficient Community Detection
Authors: Yun Jin Park, Didong Li
TMLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through applications on multiple real-world datasets, including the NCAA football league network, the DBLP collaboration network, the Amazon product co-purchasing network, and the You Tube social network, we demonstrate the efficacy of our method in significantly improving the performance of various community detection algorithms. |
| Researcher Affiliation | Academia | Yun Jin Park EMAIL Department of Biostatistics The University of North Carolina at Chapel Hill Didong Li EMAIL Department of Biostatistics The University of North Carolina at Chapel Hill |
| Pseudocode | Yes | Algorithm 1: LRC-based preprocessing algorithm for community detection Input: Raw network data: G = (V, E) Output: Preprocessed network data G = (V, E ) 1 Calculate the LRC for all edges; 2 Fit a Gaussian mixture model with two mixing component to LRCs, obtaining the estimate ˆp(x) = π1N(x; µ1, σ2 1) + π2N(x; µ2, σ2 2), where µ1 < µ2; 3 Find the local minimum β := inf µ1<x<µ2 ˆp(x); 4 Remove all edges with LRCs smaller than β: E := {(ij) E : LRC(ij) β} |
| Open Source Code | Yes | All codes can be found in https://github.com/parkyunjin/Lower Ricci Curv |
| Open Datasets | Yes | The four real datasets used in this paper can be downloaded in the following websites: 1. NCAA Football League network: https://websites.umich.edu/~mejn/netdata/ under American College football". The graph is provided in the Graph Modeling Language (.gml) format. 2. DBLP collaboration network: https://snap.stanford.edu/data/com-DBLP.html This graph is represented as Network X objects, provided by the CDlib Python package. 3. Amazon product co-purchasing network: https://snap.stanford.edu/data/com-Amazon.html This graph is represented as Network X objects, provided by the CDlib Python package. 4. You Tube social network: https://snap.stanford.edu/data/com-Youtube.html This graph is represented as Network X objects, provided by the CDlib Python package. |
| Dataset Splits | No | The paper describes using real-world datasets with "known community structures" or "ground-truth conference groups" for evaluation, but does not specify explicit training/test/validation splits for the datasets, as the community detection algorithms typically operate on the entire network for evaluation against ground truth. |
| Hardware Specification | No | The paper mentions "runtime (in seconds)" for algorithms in tables but provides no specific details about the hardware (CPU, GPU, memory, etc.) used for these computations. |
| Software Dependencies | No | All algorithms implemented in this paper are from Python package CDlib (Rossetti et al., 2019). This mentions a package but without a specific version number. No other software dependencies are listed with version numbers. |
| Experiment Setup | Yes | E.1 Hyperparameters for community detection All algorithms implemented in this paper are from Python package CDlib (Rossetti et al., 2019). The hyperparameters are as follows: 1. NCAA Football League network Label Propagation: NA. Leiden: Initial membership = None, weights= None. Girvan-Newman: Level = 10. Walktrap: NA. 2. DBLP collaboration network Angel: Threshold = 0.5, minimum community size = 3. Ego-networks: Level = 1. K-clique: K = 3. SLPA: t = 20, r = 0.1. [Similar details are provided for Amazon and You Tube networks]. |