Research map / Graph learning / Knowledge and temporal graphs
Adaptive Interaction Modeling via Graph Operations Search
Haoxin Li, Wei-Shi Zheng, Yu Tao, Haifeng Hu, Jian-Huang Lai
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 searched adaptive structures rather than global pooling | I3D-ResNet interaction recognition · Something-Something-V1 Val | accuracy: 48.1 -> 51.4 | Improved |
| Use searched adaptive structures rather than global pooling | I3D-ResNet interaction recognition · Something-Something-V2 Val | accuracy: 60.3 -> 63.5 | Improved |
| Use an adaptive structure per video rather than a single non-adaptive structure at test time (ablation) | Graph operations search · Something-Something-V1/V2 Val | accuracy: 50.2 -> 51.4; 62.4 -> 63.5 | Improved |
| Add feature aggregation as the graph operation (ablation) | I3D-ResNet interaction recognition · Something-Something-V1/V2 Val | accuracy: 48.1 -> 49.9; 60.3 -> 62.0 | Improved |
| Add difference propagation as the graph operation (ablation) | I3D-ResNet interaction recognition · Something-Something-V1/V2 Val | accuracy: 48.1 -> 49.5; 60.3 -> 61.8 | Improved |
| Add temporal convolution as the graph operation (ablation) | I3D-ResNet interaction recognition · Something-Something-V1/V2 Val | accuracy: 48.1 -> 48.7; 60.3 -> 61.0 | Improved |
| Add background incorporation as the graph operation (ablation) | I3D-ResNet interaction recognition · Something-Something-V1/V2 Val | accuracy: 48.1 -> 49.7; 60.3 -> 62.4 | Improved |
| Add node attention as the graph operation (ablation) | I3D-ResNet interaction recognition · Something-Something-V1/V2 Val | accuracy: 48.1 -> 49.8; 60.3 -> 61.8 | Improved |
| Use fixed substructures instead of original graph operations as search primitives (ablation) | Graph operations search · Something-Something-V1/V2 Val | accuracy: 51.0 -> 51.4; 63.1 -> 63.5 | Improved |
| Add diversity regularization (ablation) | Non-adaptive graph operations search · Something-Something-V1 Val | accuracy: 49.9 -> 50.7 | Improved |
| Transfer learned adaptive structure weights across datasets (ablation) | Adaptive structures search · Something-Something-V1/V2 Val | accuracy: 51.4 -> 51.0; 63.5 -> 63.4 | No clear effect |
| Increase the computation cell from 2 to 3 intermediate supernodes (ablation) | Adaptive structures search · Something-Something-V1/V2 Val | accuracy: 50.6 -> 51.4; 63.1 -> 63.5 | Improved |
Closest papers by meaning
- Learning Asynchronous and Sparse Human-Object Interaction in Videos (CVPR 2021 · similarity 0.88)
- Anticipating Future Relations via Graph Growing for Action Prediction (AAAI 2021 · similarity 0.85)
- EAN: Event Adaptive Network for Enhanced Action Recognition (IJCV 2022 · similarity 0.84)
- GraphVid: Interactive Graph-Controllable Video Generation (ECCV 2026 · similarity 0.84)
- GraphMimic: Graph-to-Graphs Generative Modeling from Videos for Policy Learning (CVPR 2025 · similarity 0.84)
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · 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.
