Research map / Graph learning / GNN architectures
Graph Structure of Neural Networks
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 |
|---|---|---|---|
| Vary MLP hidden width from 512 to 256 or 1024 (ablation) | 5-layer MLP · CIFAR-10 | correlation of graph performance across widths | No clear effect |
| Increase MLP depth from 5 to 7 layers (ablation) | 512-dim MLP · CIFAR-10 | correlation of graph performance across depths | No clear effect |
| Decrease MLP depth from 5 to 3 layers (ablation) | 512-dim MLP · CIFAR-10 | correlation of graph performance across depths | Mixed |
| Vary relational graph node count from 64 to 52 or 71 (ablation) | 5-layer MLP · CIFAR-10 | predictive performance | No clear effect |
| Use a complete graph instead of sparse relational graphs (ablation) | 8-layer 64-dim CNN · CIFAR-10 | predictive performance | Helped |
| Reduce CNN width from 64 to 16 dimensions, using 16-node relational graphs (ablation) | 8-layer CNN · CIFAR-10 | predictive performance | Mixed |
| Use sparse relational graphs in the reported ImageNet experiments rather than a complete graph | neural network models · ImageNet | predictive performance | Helped |
| Use sparse relational graphs in the reported CNN experiments rather than a complete graph (ablation) | CNN models · CIFAR-10 | predictive performance | Mixed |
| Vary graph structure measures across relational graphs | 5-layer MLP · CIFAR-10 | predictive performance | Mixed |
Closest papers by meaning
- On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks (ICML 2025 · similarity 0.89)
- Learning Latent Graph Structures and their Uncertainty (ICML 2025 · similarity 0.89)
- NN-Former: Rethinking Graph Structure in Neural Architecture Representation (CVPR 2025 · similarity 0.87)
- What do Graph Neural Networks learn? Insights from Tropical Geometry (NeurIPS 2024 · similarity 0.87)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.87)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · 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.