Research map / Probabilistic and scientific modeling / Molecules and proteins
CAT-Walk: Inductive Hypergraph Learning via Set Walks
Ali Behrouz, Farnoosh Hashemi, Sadaf Sadeghian, Margo Seltzer
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 CAT-Walk rather than the strongest listed baseline | Hyperedge prediction · Strongly inductive: all eight Table 1 datasets | AUC (%) | Improved |
| Use CAT-Walk rather than the strongest listed baseline | Hyperedge prediction · Transductive: seven datasets other than Email Enron | AUC (%) | Improved |
| Use CAT-Walk rather than the strongest listed baseline | Hyperedge prediction · Email Enron, transductive | AUC (%): 83.59 ± 0.98 vs 80.45 ± 5.30 | Hurt |
| Add SetWalk instead of using a random walk (ablation) | CAT-Walk · High School, inductive | AUC: 92.10 ± 2.18 -> 96.03 ± 1.50 | Improved |
| Add SetWalk instead of using a random walk (ablation) | CAT-Walk · Primary School, inductive | AUC: 51.56 ± 5.63 -> 95.32 ± 0.89 | Improved |
| Add time encoding (ablation) | CAT-Walk · Users in Threads, inductive | AUC: 70.58 ± 9.32 -> 89.84 ± 6.02 | Improved |
| Use SetMixer instead of Mean pooling (ablation) | CAT-Walk · High School, inductive | AUC: 94.58 ± 1.22 -> 96.03 ± 1.50 | Improved |
| Use SetMixer instead of a sum-based universal set approximator (ablation) | CAT-Walk · Question Tags U, inductive | AUC: 89.76 ± 0.45 -> 97.59 ± 2.21 | Improved |
| Use MLP-Mixer instead of RNN for walk encoding (ablation) | CAT-Walk · Primary School, inductive | AUC: 50.29 ± 4.07 -> 95.32 ± 0.89 | Improved |
| Use nonzero temporal bias rather than fixing alpha = 0 (ablation) | CAT-Walk · High School, inductive | AUC: 74.06 ± 14.9 -> 96.03 ± 1.50 | Improved |
| Add time encoding (ablation) | CAT-Walk · Congress Bill, inductive | AUC: 92.56 ± 0.49 -> 93.54 ± 0.56 | Improved |
Closest papers by meaning
- SPHINX: Structural Prediction using Hypergraph Inference Network (ICML 2025 · similarity 0.89)
- Defining and Discovering Hyper-meta-paths for Heterogeneous Hypergraphs (NeurIPS 2025 · similarity 0.89)
- Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks (ICLR 2021 · similarity 0.89)
- HyperMixup: Hypergraph-Augmented with Higher-order Information Mixup (NeurIPS 2025 · similarity 0.89)
- Deep Representation Learning for Forecasting Recursive and Multi-Relational Events in Temporal Networks (AAAI 2025 · similarity 0.87)
- Training-Free Message Passing for Learning on Hypergraphs (ICLR 2025 · 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.
