Research map / Graph learning / GNN architectures
Beyond GNNs: An Efficient Architecture for Graph Problems
Pranjal Awasthi, Abhimanyu Das, Sreenivas Gollapudi
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 GNN+ instead of message-passing GNNmp | Graph problem predictors · Shortest Paths | test MSE: 0.975 ± 0.031 -> 0.849 ± 0.022 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph problem predictors · Effective Resistance | test MSE: 0.397 ± 0.025 -> 0.187 ± 0.008 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph problem predictors · Affinity | test MSE: 0.0025 ± 0.00017 -> 0.0018 ± 1.8e−05 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph problem predictors · MST | test MSE: 1011.39 ± 106.94 -> 733.901 ± 30.97 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph problem predictors · MINCUT | test MSE: 0.963 ± 0.110 -> 0.694 ± 0.07 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph classification · IMDB-BINARY | test accuracy: 0.742 ± 0.09 -> 0.769 ± 0.02 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph classification · IMDB-MULTI | test accuracy: 0.523 ± 0.06 -> 0.527 ± 0.04 | No clear effect |
| Use GNN+ instead of message-passing GNNmp | Graph classification · COLLAB | test accuracy: 0.802 ± 0.02 -> 0.816 ± 0.004 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph classification · PROTEINS | test accuracy: 0.760 ± 0.008 -> 0.765 ± 0.015 | No clear effect |
| Use GNN+ instead of message-passing GNNmp | Graph classification · NCI1 | test accuracy: 0.849 ± 0.004 -> 0.851 ± 0.003 | No clear effect |
| Use GNN+ instead of message-passing GNNmp | Graph classification · PTC | test accuracy: 0.686 ± 0.02 -> 0.708 ± 0.018 | Improved |
| Use GNN+ instead of message-passing GNNmp | Graph classification · MUTAG | test accuracy: 0.876 ± 0.016 -> 0.898 ± 0.012 | Improved |
Closest papers by meaning
- Parallel and Distributed Graph Neural Networks: An In-Depth Concurrency Analysis (TPAMI 2024 · similarity 0.90)
- Neural Execution of Graph Algorithms (ICLR 2020 · similarity 0.89)
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.89)
- Geometry-Aware Edge Pooling for Graph Neural Networks (NeurIPS 2025 · similarity 0.88)
- On the Power of Small-size Graph Neural Networks for Linear Programming (NeurIPS 2024 · similarity 0.88)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.87)
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.
