Stability of Graph Scattering Transforms

Authors: Fernando Gama, Alejandro Ribeiro, Joan Bruna

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we show through numerical experiments in section 5, that the GST representation is not only stable, but also captures rich enough information.
Researcher Affiliation Academia Fernando Gama Dept. of Electrical and Systems Eng. University of Pennsylvania Philadelphia, PA 19104 fgama@seas.upenn.edu Joan Bruna Courant Institute of Math. Sci. New York University New York, NY 10012 bruna@cims.nyu.edu Alejandro Ribeiro Dept. of Electrical and Systems Eng. University of Pennsylvania Philadelphia, PA 19104 aribeiro@seas.upenn.edu
Pseudocode No The paper describes the architecture and computations mathematically but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes Datasets and source code: http://github.com/alelab-upenn/graph-scattering-transforms
Open Datasets Yes For the second and third experiments, we consider two problems involving real-world data... authorship attribution and source localization over a Facebook subgraph, namely the same problems considered in [22]. The references [37], [38], and [39] point to the specific datasets and problems used.
Dataset Splits No The paper refers to using datasets for 'authorship attribution' and 'source localization' problems, noting they are 'in the same scenario considered in [22]'. However, this paper does not explicitly provide specific train/validation/test split percentages, sample counts, or a detailed splitting methodology for reproducibility.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory, or specific computing cluster types) used to run the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x, or specific library versions) that would be needed to replicate the experiment environment.
Experiment Setup Yes We consider GSTs with 6 scales and 3 layers, yielding representations with 43 coefficients when using the monic cubic polynomial [31] and a tight Hann wavelet [32]. For the geometric scattering we consider the low pass operator to compute 4 moments, as used in [28], leading to 172 coefficients.