Research map / Graph learning / GNN architectures
CNN2GNN: How to Bridge CNN With GNN
Ziheng Jiao, Hongyuan Zhang, Xuelong Li
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add differentiable sparse graph head (vs inner-product graph) (ablation) | GNN student · CIFAR-100 | accuracy (%): 7.82 -> 59.77 | Improved |
| add differentiable sparse graph head (vs Euclidean graph) (ablation) | GNN student · CIFAR-100 | accuracy (%): 6.15 -> 59.77 | Improved |
| add differentiable sparse graph head (vs inner-product graph) (ablation) | GNN student · Mini-ImageNet | accuracy (%): 6.17 -> 49.07 | Improved |
| add differentiable sparse graph head (vs Euclidean graph) (ablation) | GNN student · Mini-ImageNet | accuracy (%): 5.82 -> 48.85 | Improved |
| use batch-by-batch instead of one-by-one inference (ablation) | CNN2GNN · CIFAR-100 | inference time (s): 4999.39 -> 2797.42 | Improved |
| use batch-by-batch instead of one-by-one inference (ablation) | CNN2GNN · CIFAR-100 | accuracy (%): 59.77 -> 58.57 | Hurt |
| use batch-by-batch instead of one-by-one inference (ablation) | CNN2GNN · Mini-ImageNet | accuracy (%): 48.85 -> 49.07 | No clear effect |
| add heterogeneous CNN-to-GNN distillation to GNN (vs plain GNN) | GNN student ResNet34->ResNet18 · Mini-ImageNet | accuracy (%): 44.27 -> 57.94 | Improved |
| add heterogeneous CNN-to-GNN distillation to GNN (vs plain GNN) | GNN student ResNet34->ResNet18 · CIFAR-100 | accuracy (%): 60.21 -> 65.13 | Improved |
| replace KD with CNN2GNN | student · Mini-ImageNet ResNet34->ResNet18 | accuracy (%): 45.07 -> 57.94 | Improved |
| increase temperature tau (ablation) | CNN2GNN · CIFAR-100 | accuracy | Improved |
| increase sparsity s (ablation) | CNN2GNN · Mini-ImageNet | accuracy | Improved |
Closest papers by meaning
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.88)
- NN-Former: Rethinking Graph Structure in Neural Architecture Representation (CVPR 2025 · similarity 0.87)
- MuseGNN: Forming Scalable, Convergent GNN Layers that Minimize a Sampling-Based Energy (ICLR 2025 · similarity 0.87)
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective (NeurIPS 2025 · similarity 0.87)
- TINED: GNNs-to-MLPs by Teacher Injection and Dirichlet Energy Distillation (ICML 2025 · similarity 0.87)
- Diffusing to the Top: Boost Graph Neural Networks with Minimal Hyperparameter Tuning (ICLR 2025 · similarity 0.87)
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.
