Research map / Graph learning / Graph self-supervised learning
Graph Mixup with Soft Alignments
Hongyi Ling, Zhimeng Jiang, Meng Liu, Shuiwang Ji, Na Zou
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 S-Mixup soft-alignment mixup vs no augmentation | GCN · REDDIT-BINARY | accuracy (%): +4.45 | Improved |
| add S-Mixup soft-alignment mixup vs no augmentation | GIN · six TUDatasets and ogbg-molhiv | accuracy / ROC-AUC: up to +2.25 on IMDB-B, +2.12 on NCI1 (GIN) | Improved |
| random node-order mixing (ifMixup-style) instead of soft alignment (ablation) | GIN · MOTIF | accuracy (%): 91.47 -> 52.88 | Hurt |
| mix only same-class graphs instead of all classes (ablation) | S-Mixup · IMDB-BINARY, REDDIT-BINARY, IMDB-MULTI | accuracy (%): IMDB-B 74.40->73.07; REDDIT-B 89.30->87.60; IMDB-M 50.73->50.27 | Hurt |
| widen mixup ratio lambda range to [0,1] instead of [0.5,1] (ablation) | S-Mixup · IMDB-BINARY, REDDIT-BINARY, IMDB-MULTI | accuracy (%): IMDB-B 74.40->73.00; REDDIT-B 89.30->88.70; IMDB-M 50.73->49.07 | Hurt |
| replace learned graph matching with RRWM traditional matching (ablation) | S-Mixup (GIN) · IMDB-BINARY, IMDB-MULTI | accuracy (%): IMDB-B 73.40->72.80; IMDB-M 50.13->50.11 | No clear effect |
| replace softmax normalization with Sinkhorn normalization in soft alignment (ablation) | S-Mixup (GCN) · IMDB-BINARY, REDDIT-BINARY, NCI1, IMDB-MULTI | accuracy (%): IMDB-B 74.40->73.20; REDDIT-B 89.30->90.50; NCI1 75.47->74.89 | No clear effect |
| increase Beta alpha from 0.5 to 5 (ablation) | S-Mixup (GCN) · REDDIT-BINARY | accuracy (%): 89.30 -> 77.25 | Hurt |
| add S-Mixup under 40-60% label corruption vs vanilla | GCN · IMDB-BINARY and IMDB-MULTI with noisy labels | accuracy (%): IMDB-B 40%: 62.70->65.10; 60%: 45.80->48.90 | Improved |
| add S-Mixup vs graph-level M-Mixup and SubMix | GIN · PROTEINS | accuracy (%): S-Mixup 69.37 vs Vanilla 68.28 but below DropNode 69.64 and SubMix-like baselines | Mixed |
Closest papers by meaning
- AGMixup: Adaptive Graph Mixup for Semi-supervised Node Classification (AAAI 2025 · similarity 0.90)
- Graph Invariant Learning with Subgraph Co-mixup for Out-of-Distribution Generalization (AAAI 2024 · similarity 0.89)
- IntraMix: Intra-Class Mixup Generation for Accurate Labels and Neighbors (NeurIPS 2024 · similarity 0.89)
- NodeMixup: Tackling Under-Reaching for Graph Neural Networks (AAAI 2024 · similarity 0.88)
- Graph Mixup on Approximate Gromov–Wasserstein Geodesics (ICML 2024 · similarity 0.86)
- Pairwise Alignment Improves Graph Domain Adaptation (ICML 2024 · similarity 0.86)
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.
