Research map / Graph learning / GNN architectures
DeepSphere: a graph-based spherical CNN
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Increase graph neighbors from 8 to 20 (ablation) | DeepSphere · Cosmological model classification | accuracy: 87.1 -> 91.3 | Helped |
| Increase graph neighbors from 20 to 40 (ablation) | DeepSphere · Cosmological model classification | accuracy: 91.3 -> 92.5 | Helped |
| Use HEALPix instead of equiangular sampling (ablation) | DeepSphere · SHREC'17 | mAP: 66.5 -> 68.6 | Helped |
| Use weighted rather than non-weighted loss (ablation) | DeepSphere · Climate event segmentation | mean accuracy: $87.8\pm 0.5$ -> $97.8\pm 0.3$ | Helped |
| Use weighted rather than non-weighted loss (ablation) | DeepSphere · Climate event segmentation | mean mAP: $89.16\pm 1.37$ -> $77.15\pm 1.94$ | Hurt |
| Use the wider architecture instead of the original architecture (ablation) | DeepSphere with non-weighted loss · Climate event segmentation | mean mAP: $89.16\pm 1.37$ -> 91.38 | Helped |
| Use the wider architecture instead of the original architecture (ablation) | DeepSphere with weighted loss · Climate event segmentation | mean accuracy: $97.8\pm 0.3$ -> 94.6 | Hurt |
| Increase polynomial order from 0 to 4 to incorporate neighborhood structure (ablation) | DeepSphere · Weather station regression: temperature from past temperature | MSE: 10.88 -> 8.20 | Helped |
| Increase polynomial order from 0 to 4 to incorporate neighborhood structure (ablation) | DeepSphere · Weather station regression: day from temperature | MSE: 0.10 -> 0.05 | Helped |
| Increase polynomial order from 0 to 4 to incorporate neighborhood structure (ablation) | DeepSphere · Weather station regression: day from precipitation | MSE: 0.58 -> 0.50 | Helped |
| Use the optimal kernel width instead of the prior weighting scheme's width (ablation) | DeepSphere with k=8 · Rotation equivariance evaluation | equivariance error | Helped |
| Increase the number of graph neighbors (ablation) | DeepSphere · Rotation equivariance evaluation | equivariance error | Helped |
Closest papers by meaning
- Efficient Generalized Spherical CNNs (ICLR 2021 · similarity 0.92)
- Scalable and Equivariant Spherical CNNs by Discrete-Continuous (DISCO) Convolutions (ICLR 2023 · similarity 0.91)
- Rotation Equivariant Graph Convolutional Network for Spherical Image Classification (CVPR 2020 · similarity 0.91)
- Equivariance versus Augmentation for Spherical Images (ICML 2022 · similarity 0.89)
- Spherical CNNs (ICLR 2018 · similarity 0.89)
- O$n$ Learning Deep O($n$)-Equivariant Hyperspheres (ICML 2024 · similarity 0.86)
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.