Research map / Graph learning / Graph algorithms and clustering
Set2Graph: Learning Graphs From Sets
Hadar Serviansky, Nimrod Segol, Jonathan Shlomi, Kyle Cranmer, Eilam Gross, Haggai Maron, Yaron Lipman
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 |
|---|---|---|---|
| Use global set context rather than independent Siamese element embeddings (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.605 -> 0.646 | Improved |
| Use global set context rather than independent Siamese element embeddings (ablation) | S2G particle partitioning · Charm-jets | F1: 0.728 -> 0.750 | Improved |
| Use global set context rather than independent Siamese element embeddings (ablation) | S2G particle partitioning · Light-jets | F1: 0.973 -> 0.972 | No clear effect |
| Add attention to the DeepSets set encoder (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.640 -> 0.646 | Improved |
| Add attention to the DeepSets set encoder (ablation) | S2G particle partitioning · Charm-jets | F1: 0.743 -> 0.750 | Improved |
| Add one set-encoder layer (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.649 -> 0.660 | Improved |
| Use DeepSets instead of PointNetSeg as the set encoder (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.630 -> 0.646 | Improved |
| Use two edge-network layers instead of one (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.565 -> 0.646 | Improved |
| Use full equivariant-basis broadcasting instead of basic broadcasting (ablation) | S2G particle partitioning · Bottom-jets | F1: 0.646 -> 0.655 | No clear effect |
| Use global set reasoning rather than pairwise Siamese learning | S2G · Delaunay, n=50 | F1: 0.704 -> 0.926 | Improved |
| Use global set reasoning rather than pairwise Siamese learning | S2G · Convex hull, Gaussian, 30 points | F1: 0.275 -> 0.707 | Improved |
| Apply S2G instead of GNN5 | Convex-hull prediction · Spherical, 20-100 points | F1: 0.667 -> 0.535 | Hurt |
Closest papers by meaning
- You are AllSet: A Multiset Function Framework for Hypergraph Neural Networks (ICLR 2022 · similarity 0.79)
- Hypergraph Joint Representation Learning for Hypervertices and Hyperedges via Cross Expansion (AAAI 2024 · similarity 0.78)
- Hypergraph Learning for Unsupervised Graph Alignment via Optimal Transport (AAAI 2025 · similarity 0.76)
- Hypergraph-enhanced Dual Semi-supervised Graph Classification (ICML 2024 · similarity 0.76)
- Training-Free Message Passing for Learning on Hypergraphs (ICLR 2025 · similarity 0.76)
- SPHINX: Structural Prediction using Hypergraph Inference Network (ICML 2025 · similarity 0.76)
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.
