Adversarial Permutation Guided Node Representations for Link Prediction

Authors: Indradyumna Roy, Abir De, Soumen Chakrabarti9445-9453

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on diverse datasets show that PERMGNN outperforms several state-of-the-art link predictors, and can predict the most likely edges fast.
Researcher Affiliation Academia Indradyumna Roy, Abir De, Soumen Chakrabarti Indian Institute of Technology Bombay {indraroy15, abir, soumen}@cse.iitb.ac.in
Pseudocode Yes Algorithm 1: Reporting ranked list of potential edges fast.
Open Source Code Yes 1Code: https://www.cse.iitb.ac.in/ abir/codes/permgnn.zip.
Open Datasets Yes Datasets. We consider five real world datasets: (1) Twitter (Leskovec and Mcauley 2012), (2) Google+ (Leskovec et al. 2010), (3) Cora (Getoor 2005; Sen et al. 2008), (4) Citeseer (Getoor 2005; Sen et al. 2008) and (5) PB (Ackland et al. 2005).
Dataset Splits Yes Then, for each q Q, in the original graph, we partition the neighbors nbr(q) and the non-neighbors nbr(q) which are within 2-hop distance from q into 70% training, 10% validation and 20% test sets, where the node pairs are sampled uniformly at random.
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models, or cloud computing instance types used for experiments.
Software Dependencies No The paper mentions “Torch” and “numpy” in the context of vectorized similarity computation but does not provide specific version numbers for these or other software dependencies.
Experiment Setup No The paper describes the evaluation protocol and dataset splits but does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) or optimizer settings.