HyperDefender: A Robust Framework for Hyperbolic GNNs

Authors: Nikita Malik, Rahul Gupta, Sandeep Kumar

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate that Hyper Defender significantly improves node classification accuracy across various attacks, effectively mitigating the performance degradation typically observed in Hy-GNNs when the hierarchy in original datasets is compromised. ... Section 4. Experiments
Researcher Affiliation Academia 1Bharti School of Telecommunication Technology and Management, Indian Institute of Technology, Delhi, India 2Department of Mathematics and Computing, Indian Institute of Technology, Delhi, India 3Department of Electrical Engineering, Indian Institute of Technology, Delhi, India 4Yardi School of Artificial Intelligence, Indian Institute of Technology, Delhi, India EMAIL, EMAIL, EMAIL
Pseudocode Yes Algorithm 1: Hyper Defender Framework ... Algorithm 2: Pro-Hy La
Open Source Code Yes Code https://github.com/nikimal99/Hyper Defender.git
Open Datasets Yes We evaluate our proposed approach on three diverse datasets: Cora (Sen et al. 2008), Disease (Anderson and May 1991) and Airport (Zhang and Chen 2018). Dataset statistics are summarized in Table 1.
Dataset Splits No The paper does not explicitly provide specific training/validation/test dataset splits (e.g., percentages, sample counts, or citations to predefined splits) needed to reproduce the experiment.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, PyTorch 1.9) needed to replicate the experiment.
Experiment Setup No Algorithm 2 lists "Hyper-parameters α, β, γ, λ, τ, Learning rate η, η , Hy La feature dimension d1" as inputs, but the paper does not provide concrete values for these hyperparameters or other system-level training settings in the main text.