Learnable Spectral Wavelets on Dynamic Graphs to Capture Global Interactions

Authors: Anson Bastos, Abhishek Nadgeri, Kuldeep Singh, Toyotaro Suzumura, Manish Singh

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on eight standard datasets show that our method significantly outperforms related methods on various tasks for dynamic graphs.
Researcher Affiliation Collaboration 1 Indian Institute of Technology Hyderabad, India 2 RWTH Aachen, Germany 3 Zerotha Research and Cerence Gmbh, Germany 4 The University of Tokyo, Japan
Pseudocode Yes 1: function fct = ESpectral(At, H(l) t , W (l) t 1) 2: W (l) t = RNN(W (l) t 1) 3: H(l+1) t = GNN(At, H(l) t , W (l) t ) 4: vgt = Pool(At, H(l+1) t ) 5: fct = W2 σ (W1vgt) 6: end function
Open Source Code No The paper does not contain any explicit statement about providing open-source code for the methodology or a link to a code repository.
Open Datasets Yes We borrow datasets and its preprocessing/splitting settings used in previous best baselines (Xiang, Huang, and Wang 2022; Pareja et al. 2020). Datasets: Table ?? summarizes eight datasets for link prediction, edge classification, and node classification. Each dataset contains a sequence of time-ordered graphs. ...Finally, for Brain dataset, nodes represent tiny regions/cubes in the brain, and the edges are their connectivity. (Xu et al. 2019b)
Dataset Splits Yes Datasets: Table ?? summarizes eight datasets for link prediction, edge classification, and node classification. Each dataset contains a sequence of time-ordered graphs. ... # Time Steps Task (Train / Val / Test) BC-OTC 95 / 14 / 28 EC
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, or memory) used to run the experiments.
Software Dependencies No The paper does not provide specific version numbers for any software components or libraries used in the experiments.
Experiment Setup No The best hyperparameters search has the range as: Number of layers {1, 2}, Hidden dimension {32, 64, 128}, Number of heads {4, 8, 16}, Filter order {4, 8, 16}, Wavelet scales [0.1, 10]. The rest of the parameters and settings are borrowed from previous works (Pareja et al. 2020; Xiang, Huang, and Wang 2022).