Research map / Graph learning / Graph self-supervised learning
Graph Convolution for Semi-Supervised Classification: Improved Linear Separability and Out-of-Distribution Generalization
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 |
|---|---|---|---|
| Add graph convolution before linear classification | Semi-supervised CSBM classification · CSBM with class-mean distance above the convolution separability threshold | linear separability: P((\tilde{X}_i)_{i\in S} is linearly separable) = 1-o_d(1); P((X_i)_{i\in S} is linearly separable)≤C exp(−cd) | Helped |
| Add graph convolution before linear classification | Semi-supervised CSBM classification · CSBM with high graph noise and well-separated class means | test error | Hurt |
| Increase graph density while keeping the graph signal-to-noise ratio fixed | Graph-convolution classifier · CSBM, q=0.2p | training and test loss | Helped |
| Use graph convolution on graphs with p,q ≤ log^2 n /n | CSBM classification · Sparse CSBM | training and test loss | No clear effect |
| Train with graph convolution and test on graphs with different intra- and inter-class edge probabilities | CSBM graph-convolution classifier · Out-of-distribution CSBM with unchanged attribute means | test binary cross-entropy | Helped |
| Add graph convolution | Linear classifiers on Cora, PubMed, and WikipediaNetwork · Real-data one-vs-all classification with noisy test graphs | test error | Mixed |
| Add graph convolution when the original attributes are already linearly separable | Semi-supervised classification · CSBM with widely separated class means | test error | Hurt |
Closest papers by meaning
- Analysis of Corrected Graph Convolutions (NeurIPS 2024 · similarity 0.90)
- Gaussian-Induced Convolution for Graphs (AAAI 2019 · similarity 0.90)
- Optimal Exact Recovery in Semi-Supervised Learning: A Study of Spectral Methods and Graph Convolutional Networks (ICML 2024 · similarity 0.90)
- Semi-Supervised Classification with Graph Convolutional Networks (ICLR 2017 · similarity 0.90)
- Effects of Graph Convolutions in Multi-layer Networks (ICLR 2023 · similarity 0.90)
- Unsupervised Neighborhood Propagation Kernel Layers for Semi-supervised Node Classification (AAAI 2024 · similarity 0.88)
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.