Research map / Label-efficient and robust learning / Self-supervised and contrastive
FiGURe: Simple and Efficient Unsupervised Node Representations with Filter Augmentations
Chanakya Ekbote, Ajinkya Deshpande, Arun Iyer, SUNDARARAJAN SELLAMANICKAM, Ramakrishna B Bairi
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 Random Fourier Feature projection at 32 dims (ablation) | GRACE · chameleon | accuracy (%): 53.25 -> 67.39 | Improved |
| add RFF projection at 32 dims (ablation) | DGI · chameleon | accuracy (%): 45.48 -> 56.01 | Improved |
| add RFF projection at 32 dims (ablation) | FiGURe · squirrel | accuracy (%): 39.38 -> 48.89 | Improved |
| add RFF projection at 32 dims (ablation) | FiGURe · cora | accuracy (%): 82.88 -> 82.56 | No clear effect |
| add RFF projection at 32 dims (ablation) | SUGRL · cora | accuracy (%): 65.35 -> 70.06 | Improved |
| add RFF projection at 128 dims (ablation) | MVGRL · cora | accuracy (%): 85.31 -> 84.61 | No clear effect |
| share encoder weights across filters instead of independent encoders (ablation) | FiGURe · squirrel | accuracy (%): 50.52 -> 52.23 | Improved |
| share encoder weights across filters (ablation) | FiGURe · citeseer | accuracy (%): 75.03 -> 74.77 | No clear effect |
| use 2-layer GCN encoder instead of 1 layer (ablation) | FiGURe · chameleon | accuracy (%): 68.55 -> 53.53 | Hurt |
| use ANOVA kernel random features instead of RFF (ablation) | FiGURe · cora | accuracy (%): 87.00 -> 83.26 | Hurt |
| use ChebNet K=11 filter bank instead of GPRGNN filters (ablation) | FiGURe · cora | accuracy (%): 87.00 -> 76.14 | Hurt |
| use BernNet K=11 filters instead of K=3 (ablation) | FiGURe · squirrel | accuracy (%): 37.07 -> 43.48 | Improved |
Closest papers by meaning
- Beyond Supervised vs. Unsupervised: Representative Benchmarking and Analysis of Image Representation Learning (CVPR 2022 · similarity 0.87)
- URLOST: Unsupervised Representation Learning without Stationarity or Topology (ICLR 2025 · similarity 0.86)
- From $t$-SNE to UMAP with contrastive learning (ICLR 2023 · similarity 0.85)
- Holographic Node Representations: Pre-training Task-Agnostic Node Embeddings (ICLR 2025 · similarity 0.85)
- $\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs (ICLR 2025 · similarity 0.85)
- Towards an Explainable Comparison and Alignment of Feature Embeddings (ICML 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.
