FUGAL: Feature-fortified Unrestricted Graph Alignment

Authors: Aditya Bommakanti, Harshith Vonteri, Konstantinos Skitsas, Sayan Ranu, Davide Mottin, Panagiotis Karras

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimentation demonstrates that FUGAL consistently surpasses state-of-the-art graph alignment methods in accuracy across all benchmark datasets without encumbering efficiency.
Researcher Affiliation Academia Aditya Bommakanti IIT Delhi adityabommakanti2002@gmail.com Harshith Reddy Vonteri IIT Delhi harshithreddyvonteri@gmail.com Konstantinos Skitsas Aarhus University skitsas@cs.au.dk Sayan Ranu IIT Delhi sayanranu@cse.iitd.ac.in Davide Mottin Aarhus University davide@cs.au.dk Panagiotis Karras University of Copenhagen & Aarhus University piekarras@gmail.com
Pseudocode Yes Algorithm 1 FINDQUASIPERMUTATION (A, B, D, µ, T) ... Alg. 2 in the appendix presents the complete FUGAL pseudocode.
Open Source Code Yes 1Code and data at https://github.com/idea-iitd/Fugal.
Open Datasets Yes Table 1: Real-graph nodes n, edges m, and network type. ... Synthetic Graphs. We employ Newmann-Watts (NW) [21] graphs, characterized by small-world properties and a high clustering coefficient.
Dataset Splits No The paper discusses evaluation on noisy variants and different graph versions (e.g., 'align the last graph version to versions containing 80%, 85%, 90%, and 99% of edges') but does not explicitly provide traditional training/test/validation dataset splits (e.g., percentages or sample counts) needed to reproduce data partitioning for model training or selection.
Hardware Specification Yes We ran all experiments on a 40-core Intel Xeon E5-2687W CPU machine @3.10GHz with Python implementations of FUGAL1 and competitors;
Software Dependencies No The paper mentions 'Python implementations' of their method and competitors but does not provide specific version numbers for Python or any other key software libraries or dependencies.
Experiment Setup Yes Table 3 in Appendix lists the parameters we employ in FUGAL with each dataset. We set the number of iterations T to 15 for all datasets. The parameter µ controls the sway of node features in the optimization. ... We employ two noise types: one-way noise removes edges from the target graph, while bimodal noise removes and restores the same number of edges.