Research map / Graph learning / Knowledge and temporal graphs
Interpretable Graph Capsule Networks for Object Recognition
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 |
|---|---|---|---|
| Replace capsule routing with multi-head attention-based graph pooling | CapsNets · CIFAR10 | accuracy: 74.64{ ($\pm$ 1.02)} -> 82.21 | Helped |
| Replace capsule routing with multi-head attention-based graph pooling | CapsNets · MNIST | accuracy: 99.41{ ($\pm$ 0.08)} -> 99.50 | Helped |
| Replace capsule routing with multi-head attention-based graph pooling | CapsNets · Fashion MNIST | accuracy: 92.12{ ($\pm$ 0.29)} -> 93.1 | Helped |
| Use multi-head graph pooling instead of routing | ResNet20 capsule models · CIFAR10 | error rate: 8.46($\pm$0.27) -> 7.99($\pm$0.13) | Helped |
| Use multi-head graph pooling instead of routing | ResNet20 capsule models · SVHN | error rate: 3.49($\pm$0.69) -> 2.98($\pm$0.09) | Helped |
| Increase the number of graph-pooling heads (ablation) | GraCapsNets · Fashion MNIST and CIFAR10 | classification performance | Mixed |
| Model spatial relationships between capsules in graph pooling (ablation) | GraCapsNets · Fashion MNIST and CIFAR10 | classification performance | Helped |
| Use top-K graph pooling without graph structure (ablation) | GraCapsNets · Fashion MNIST and CIFAR10 | classification performance | Hurt |
| Use direct trainable routing without graph information (ablation) | GraCapsNets · Fashion MNIST and CIFAR10 | classification performance | Hurt |
| Average transformed capsules without routing (ablation) | GraCapsNets · Fashion MNIST and CIFAR10 | classification performance | Hurt |
| Use built-in graph attention rather than the compared saliency methods | GraCapsNets · Fashion MNIST and CIFAR10 | AOPC | Helped |
| Apply FGSM or C&W attacks to GraCapsNets instead of the CNN and capsule baselines | GraCapsNets · SVHN and CIFAR10 | attack success rate | Helped |
Closest papers by meaning
- Explainability Methods for Graph Convolutional Neural Networks (CVPR 2019 · similarity 0.87)
- STAR-Caps: Capsule Networks with Straight-Through Attentive Routing (NeurIPS 2019 · similarity 0.87)
- Why Capsule Neural Networks Do Not Scale: Challenging the Dynamic Parse-Tree Assumption (AAAI 2023 · similarity 0.86)
- Capsule Graph Neural Network (ICLR 2019 · similarity 0.86)
- ParseCaps: An Interpretable Parsing Capsule Network for Medical Image Diagnosis (AAAI 2025 · similarity 0.85)
- Show and Tell: Visually Explainable Deep Neural Nets via Spatially-Aware Concept Bottleneck Models (CVPR 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.