Research map / Graph learning / Graph self-supervised learning
Association Graph Learning for Multi-Task Classification with Category Shifts
Jiayi Shen, Zehao Xiao, Xiantong Zhen, Cees G. M. Snoek, Marcel Worring
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 assignment entropy maximization (L_AE) (ablation) | association graph model · Office-Home, 75% missing | harmonic mean H: 58.26 -> 60.59 | Improved |
| add assignment entropy maximization (L_AE) (ablation) | association graph model · Office-Home, 75% missing | A_m: 44.65 -> 47.51 | Improved |
| add assignment entropy maximization (L_AE) (ablation) | association graph model · Office-Home, 75% missing | A_o: 87.59 -> 87.16 | No clear effect |
| add assignment entropy maximization (L_AE) (ablation) | association graph model · Office-Home, 75% missing | average assignment entropy: 0.13 -> 0.84 | Improved |
| increase neighbor size from 0 to maximum 70 (ablation) | association graph model · Office-Home, 75% missing | harmonic mean H: 49.82 -> 60.59 | Improved |
| increase neighbor size from 8 to 70 (ablation) | association graph model · Office-Home, 75% missing | harmonic mean H: 58.45 -> 60.59 | Improved |
| increase neighbor size from 0 to 70 (ablation) | association graph model · Office-Home, 75% missing | A_m: 36.84 -> 47.51 | Improved |
| add message-passing layers (L=0 to L=4) (ablation) | association graph model · Office-Home | accuracy | Improved |
| replace attention graph with association graph (type-specific metric functions) (ablation) | graph model · Office-Home | accuracy | Improved |
| use association graph instead of ERM baseline | Ours vs ERM · Office-Home, 75% missing | harmonic mean H: 49.32 -> 60.59 | Improved |
| use association graph instead of second-best WeighLosses | Ours vs WeighLosses · Office-Home, 75% missing | harmonic mean H: 50.26 -> 60.59 | Improved |
| use association graph when no categories are missing (0%) | Ours vs WeighLosses · Office-Home, 0% missing | accuracy A_o: 81.78 -> 82.01 | Improved |
Closest papers by meaning
- Exploring Correlations of Self-Supervised Tasks for Graphs (ICML 2024 · similarity 0.86)
- Distribution Matching for Multi-Task Learning of Classification Tasks: A Large-Scale Study on Faces & Beyond (AAAI 2024 · similarity 0.85)
- Towards Graph Foundation Models: Learning Generalities Across Graphs via Task-Trees (ICML 2025 · similarity 0.84)
- Replay-and-Forget-Free Graph Class-Incremental Learning: A Task Profiling and Prompting Approach (NeurIPS 2024 · similarity 0.84)
- Core Knowledge Learning Framework for Graph (AAAI 2025 · similarity 0.84)
- One For All: Towards Training One Graph Model For All Classification Tasks (ICLR 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.
