Research map / Graph learning / GNN architectures
Deep Learning with Dynamic Computation Graphs
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 |
|---|---|---|---|
| Batch operations at the same depth, including within each input tree | Tree-LSTM inference · GPU, batch-size 1 | time per tree: 0.1608 -> 0.0376 | Helped |
| Batch operations at the same depth, including within each input tree | Tree-LSTM inference · CPU, batch-size 1 | time per tree: 0.517 -> 0.258 | Helped |
| Use dynamic batching instead of manual batching | Tree-LSTM inference · GPU, batch-size 1024 | time per batch: 0.978 -> 1.590 | Hurt |
| Allow trees in the batch to have different shapes | Dynamic batching · Tree-LSTM inference | time per batch: 1.590 -> 1.617 | No clear effect |
| Ensemble thirty independently trained models | Constituency Tree-LSTM · Stanford Sentiment Treebank test, fine-grained | accuracy: 52.3 (0.7) -> 53.6 | Helped |
| Ensemble thirty independently trained models | Constituency Tree-LSTM · Stanford Sentiment Treebank test, binary | accuracy: 89.4 (0.4) -> 90.2 | Helped |
| Replace standard LSTMs with a single forget gate and pooled child histories with the Tree-LSTM variant (ablation) | Sentiment analysis models · Stanford Sentiment Treebank | accuracy | Hurt |
| Use character-level rather than word-level embeddings (ablation) | Sentiment analysis models · Stanford Sentiment Treebank | accuracy | Hurt |
Closest papers by meaning
- PRES: Toward Scalable Memory-Based Dynamic Graph Neural Networks (ICLR 2024 · similarity 0.86)
- On the Scalability of GNNs for Molecular Graphs (NeurIPS 2024 · similarity 0.85)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.85)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.85)
- Valid Conformal Prediction for Dynamic GNNs (ICLR 2025 · similarity 0.85)
- Node-Time Conditional Prompt Learning in Dynamic Graphs (ICLR 2025 · similarity 0.85)
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.