Research map / Graph learning / Graph self-supervised learning
Diffusion Mechanism in Residual Neural Network: Theory and Applications
Tangjun Wang, Zehao Dou, Chenglong Bao, Zuoqiang Shi
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 diffusion layers to ResNet (ablation) | ResNet · Cora node classification | accuracy: 58.9 -> 82.1 | Improved |
| add diffusion layers to ResNet (ablation) | ResNet · Citeseer node classification | accuracy: 61.9 -> 74.6 | Improved |
| add diffusion layers to ResNet (ablation) | ResNet · Pubmed node classification | accuracy: 70.1 -> 80.1 | Improved |
| add internal diffusion layers to convection network (ablation) | 2-layer residual network · miniImageNet 1-shot, ResNet-18 | accuracy: 53.04 -> 68.47 | Improved |
| put diffusion inside network instead of as Laplacian loss (ablation) | 2-layer residual network · miniImageNet 1-shot, ResNet-18 | accuracy: 55.30 -> 68.47 | Improved |
| put diffusion inside network instead of as Laplacian loss (ablation) | 2-layer residual network · CUB 1-shot, ResNet-18 | accuracy: 70.00 -> 79.12 | Improved |
| add internal diffusion layers (ablation) | 2-layer residual network · miniImageNet 5-shot, ResNet-18 | accuracy: 79.58 -> 80.02 | Improved |
| add internal diffusion layers (ablation) | 2-layer residual network · miniImageNet 1-shot, WRN | accuracy: 57.07 -> 69.77 | Improved |
| use Diff-ResNet instead of label-propagation Diffusion (LaplacianShot-style) (ablation) | nearest prototype features · miniImageNet 1-shot, ResNet-18 | accuracy: 57.81 -> 68.47 | Improved |
| increase network depth to 32 layers | GCN · Cora/Citeseer/Pubmed | accuracy | Hurt |
| increase network depth to 32 layers | Diff-ResNet · graph benchmarks | accuracy | Hurt |
| stack more diffusion layers at fixed total strength (ablation) | Diff-ResNet · miniImageNet 1-shot | accuracy | No clear effect |
Closest papers by meaning
- A Generalized Neural Diffusion Framework on Graphs (AAAI 2024 · similarity 0.86)
- Unlocking Dataset Distillation with Diffusion Models (NeurIPS 2025 · similarity 0.85)
- DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents (ICML 2024 · similarity 0.85)
- S4S: Solving for a Fast Diffusion Model Solver (ICML 2025 · similarity 0.85)
- Neural Flow Diffusion Models: Learnable Forward Process for Improved Diffusion Modelling (NeurIPS 2024 · similarity 0.85)
- Shallow diffusion networks provably learn hidden low-dimensional structure (ICLR 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.
