Research map / Graph learning / GNN architectures
On Dropping Clusters to Regularize Graph Convolutional Neural Networks
Xikun Zhang, Chang Xu, Dacheng Tao
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 spatial correlation (neighbor-hop dropping) to depth-wise dropping (ablation) | DropCluster · Northwestern-UCLA | Top-1 accuracy: 70.8% -> 72.2% | Improved |
| add depth-wise correlation (correlated channels) to spatial dropping (ablation) | DropCluster · Northwestern-UCLA | Top-1 accuracy: 71.1% -> 72.2% | Improved |
| add depth-wise correlation to spatial dropping (ablation) | DropCluster · Northwestern-UCLA | Top-5 accuracy: 99.1% -> 99.6% | Improved |
| replace dropout with DropCluster | ST-GCN-U · Northwestern-UCLA | Top-1 accuracy: 70.5% -> 72.2% | Improved |
| replace SpatialDropout with DropCluster | ST-GCN-U · Northwestern-UCLA | Top-1 accuracy: 70.8% -> 72.2% | Improved |
| add DropCluster vs no dropping | ST-GCN · NTU-RGB+D X-Sub | accuracy: 80.6% -> 83.3% | Improved |
| add DropCluster vs no dropping | ST-GCN · NTU-RGB+D X-View | accuracy: 88.0% -> 88.9% | Improved |
| replace Gaussian dropout with DropCluster | ST-GCN · NTU-RGB+D X-Sub | accuracy: 82.2% -> 83.3% | Improved |
| add Jumpout | ST-GCN · NTU-RGB+D X-View | accuracy: 88.0% -> 87.9% | No clear effect |
| replace dropout with DropCluster at depth 11 | GCN · NTU-RGB+D X-View (11-layer net) | accuracy: 88.6% -> 90.0% | Improved |
| add DropCluster to GECNN | GECNN · NTU-RGB+D X-Sub | accuracy: 83.6% -> 84.7% | Improved |
| add DropCluster to SLHM | SLHM · NTU-RGB+D X-Sub | accuracy: 84.3% -> 85.3% | Improved |
Closest papers by meaning
- Beyond Random Masking: When Dropout meets Graph Convolutional Networks (ICLR 2025 · similarity 0.92)
- DropEdge: Towards Deep Graph Convolutional Networks on Node Classification (ICLR 2020 · similarity 0.90)
- Decoupling GCN with DropGraph Module for Skeleton-Based Action Recognition (ECCV 2020 · similarity 0.89)
- AutoDropout: Learning Dropout Patterns to Regularize Deep Networks (AAAI 2021 · similarity 0.88)
- DropBlock: A regularization method for convolutional networks (NeurIPS 2018 · similarity 0.88)
- Effects of Dropout on Performance in Long-range Graph Learning Tasks (NeurIPS 2025 · similarity 0.88)
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.
