A Hierarchical Transitive-Aligned Graph Kernel for Un-attributed Graphs

Authors: Lu Bai, Lixin Cui, Hancock Edwin

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluations demonstrate the effectiveness of the new transitive-aligned kernel.
Researcher Affiliation Academia 1School of Artificial Intelligence, Beijing Normal University, Beijing, China. 2Central University of Finance and Economics, Beijing, China. 3Department of Computer Science, University of York, York, UK.
Pseudocode No The paper describes procedures in text and with diagrams but does not include any formal pseudocode or algorithm blocks.
Open Source Code No The paper refers to third-party implementations (K-means MATLAB, LIBSVM) but does not provide open-source code for the proposed HTAK kernel.
Open Datasets Yes We evaluate the proposed HTAK kernels on nine benchmark graph datasets from computer vision, bioinformatics, and social networks. These datasets include: BAR31, BSPHERE31, GEOD31, MUTAG, NCI1, CATH2, COLLAB, IMDB-B, and IMDB-M. ... other datasets are all available on the website http://graphkernels.cs.tu-dortmund.
Dataset Splits Yes We perform a 10-fold cross-validation where the classification accuracy is computed using a C-Support Vector Machine (C-SVM).
Hardware Specification No The paper does not provide specific details about the hardware used for experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper mentions using 'the fastest K-means MATLAB implementation developed by Deng Cai (Cai, 2012)' and 'the LIBSVM library(Chang & Lin, 2011)' but does not specify version numbers for MATLAB or LIBSVM beyond the publication year.
Experiment Setup Yes For the WLSK kernel and the JTQK kernel, we set the highest dimension ... as 10... For the ASK kernel, we set the highest layer of the required DB representation as 50... We repeat the whole experiment 10 times and report the average classification accuracy... for the proposed HTAK kernel we vary the parameter H from 1 to 5... we set the parameter Nh as Nh = 0.2Nh 1...