Limitless Stability for Graph Convolutional Networks

Authors: Christian Koke

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

Reproducibility Variable Result LLM Response
Research Type Experimental These new theoretical results are supported by corresponding numerical investigations. We focus on investigating structural perturbations, as corresponding results are most involved and novel: We first consider a graph on 5 nodes with an adjacency matrix A with Op1q-entries (c.f. 30 in Appendix N). We then scale A by 1{δa and 1{δb (with 1 δa 1 δb 1) respectively and consider the norm-difference between associated Laplacians and resolvents. Fig. 6 (a) then illustrate the theoretical result (c.f. Section 5) that resolventinstead of Laplacian-differences capture the convergence behaviour.
Researcher Affiliation Academia Anonymous authors Paper under double-blind review
Pseudocode No No pseudocode or algorithm block is explicitly labeled or provided in the paper. Figure 2 shows an "Update Rule for a GCN" as a diagram, not pseudocode.
Open Source Code No No statement regarding open-source code availability for the methodology described, nor links to repositories, are provided in the paper.
Open Datasets Yes Feature vectors are generated on the QM7 dataset. The dataset we consider is the QM7 dataset, introduced in Blum & Reymond (2009); Rupp et al. (2012).
Dataset Splits No No explicit information about training, validation, or test dataset splits (e.g., percentages, sample counts, or citations to predefined splits) is provided in the paper. The paper mentions evaluating on the QM7 dataset and averaging over "100 random unit-norm choices of f" for a specific experiment, but this is not a general dataset split.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts, or cloud resources) used for running experiments are provided in the paper.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, or other libraries) are mentioned in the paper.
Experiment Setup Yes Finally, we investigate the transferability of a two-layer GCN with 16 nodes per hidden Layer combined with the aggregation method of Section 6 into a graph-level map Ψp 2. Filters are of the form (2) up to order k 11. Coefficients tbg ku are sampled uniformly from r 100, 100s.