Research map / Graph learning / GNN architectures
Graph Scattering beyond Wavelet Shackles
Christian Koke, Gitta Kutyniok
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace graph-wavelet scattering (GS-SVM) with generalized graph scattering network (GGSN+EK) | graph scattering + SVM · REDDIT-12K | accuracy (%): 45.23 -> 49.03 | Improved |
| replace graph-wavelet scattering (GS-SVM) with GGSN+EK | graph scattering + SVM · REDDIT-B | accuracy (%): 89.65 -> 91.60 | Improved |
| replace graph-wavelet scattering (GS-SVM) with GGSN+EK | graph scattering + SVM · REDDIT-5K | accuracy (%): 53.33 -> 56.89 | Improved |
| replace graph-wavelet scattering (GS-SVM) with GGSN+EK | graph scattering + SVM · IMDB-B | accuracy (%): 71.20 -> 73.20 | Improved |
| replace graph-wavelet scattering (GS-SVM) with GGSN+EK | graph scattering + SVM · IMDB-M | accuracy (%): 48.73 -> 49.47 | Improved |
| replace graph-wavelet scattering (GS-SVM) with GGSN+EK | graph scattering + SVM · COLLAB | accuracy (%): 79.94 -> 80.34 | Improved |
| static scattering (GGSN+EK) versus fully trainable P-Poinc | GGSN+EK · IMDB-B | accuracy (%): 81.86 -> 73.20 | Hurt |
| static scattering (GGSN+EK) versus fully trainable P-Poinc | GGSN+EK · REDDIT-12K | accuracy (%): 42.16 -> 49.03 | Improved |
| static scattering (GGSN+EK) versus trainable GIN | GGSN+EK · REDDIT-B | accuracy (%): 92.40 -> 91.60 | No clear effect |
| add second-order (edge-level) scattering coefficients (GGST+EK to 2GGST+EK) (ablation) | GGST+EK · QM7 | MAE (kcal/mol): 11.3 -> 3.4 | Improved |
| 2GGST+EK versus best baseline DTNN | 2GGST+EK · QM7 | MAE (kcal/mol): 8.2 -> 3.4 | Improved |
Closest papers by meaning
- Exploitation of a Latent Mechanism in Graph Contrastive Learning: Representation Scattering (NeurIPS 2024 · similarity 0.87)
- Unsupervised Parameter-free Simplicial Representation Learning with Scattering Transforms (ICML 2024 · similarity 0.86)
- WaveNet: Tackling Non-stationary Graph Signals via Graph Spectral Wavelets (AAAI 2024 · similarity 0.86)
- Beyond Oversquashing: Understanding Signal Propagation in GNNs Via Observables (NeurIPS 2026 · similarity 0.85)
- Rayleigh Quotient Graph Neural Networks for Graph-level Anomaly Detection (ICLR 2024 · similarity 0.84)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
