Research map / Label-efficient and robust learning / Recognition and long-tail
Column Networks for Collective Classification
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 relational connections (ablation) | highway network · JBoss software delay prediction | F1: 66.8 -> 71.9 | Helped |
| use mini-batch rather than full-batch training (ablation) | Column Net-HWN · JBoss software delay prediction | F1: 71.9 -> 71.2 | Hurt |
| use a highway network rather than a standard feedforward mini-column (ablation) | Column Net · JBoss software delay prediction | F1: 70.5 -> 71.9 | Helped |
| add relational connections (ablation) | highway network · PubMed Diabetes | MicroF1-score: 87.9 -> 89.8 | Helped |
| add relational connections (ablation) | highway network · PubMed Diabetes | MacroF1-score: 87.9 -> 89.6 | Helped |
| use mini-batch rather than full-batch training (ablation) | Column Net-HWN · PubMed Diabetes | MicroF1-score: 89.8 -> 89.8 | No clear effect |
| use mini-batch rather than full-batch training (ablation) | Column Net-HWN · PubMed Diabetes | MacroF1-score: 89.6 -> 89.6 | No clear effect |
| use a highway network rather than a standard feedforward mini-column (ablation) | Column Net · PubMed Diabetes | MicroF1-score: 89.4 -> 89.8 | Helped |
| add relational connections (ablation) | highway network · MovieLens film genres | Micro-F1: 50.8 -> 57.5 | Helped |
| add relational connections (ablation) | highway network · MovieLens film genres | Macro-F1: 45.2 -> 54.1 | Helped |
| use mini-batch rather than full-batch training (ablation) | Column Net-HWN · MovieLens film genres | Macro-F1: 52.7 -> 54.1 | Helped |
| use a highway network rather than a standard feedforward mini-column (ablation) | Column Net · MovieLens film genres | Macro-F1: 41.8 -> 54.1 | Helped |
Closest papers by meaning
- Deep Collective Inference (AAAI 2017 · similarity 0.88)
- A Collective Learning Framework to Boost GNN Expressiveness for Node Classification (ICML 2021 · similarity 0.86)
- Composition-based Multi-Relational Graph Convolutional Networks (ICLR 2020 · similarity 0.84)
- Detecting Visual Relationships With Deep Relational Networks (CVPR 2017 · similarity 0.84)
- RelGNN: Composite Message Passing for Relational Deep Learning (ICML 2025 · similarity 0.83)
- Position: Relational Deep Learning - Graph Representation Learning on Relational Databases (ICML 2024 · similarity 0.83)
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.