A paper about machine learning reproducibility and the problems that arise
| .github/workflows | ||
| .idea | ||
| benchmarking/topognn | ||
| five_nodes_regression | ||
| graph_benchmark | ||
| QM9 | ||
| QM9_benchmark | ||
| report | ||
| three_nodes_classification | ||
| three_nodes_regression | ||
| tree_neighbours_match | ||
| .envrc | ||
| .gitignore | ||
| flake.lock | ||
| flake.nix | ||
| LICENSE | ||
| poetry.lock | ||
| pyproject.toml | ||
| README.md | ||
Bottlenecks within Graph Neural networks: A Study of Challenges and Mitigations
This project will explore the oversquashing problem in Graph Neural Networks (GNNs) and how it affects the ability for the model to learn long-range dependencies between nodes. The project will also explore the use of topological information in the form of the dataset's graph structure to improve the model's ability to learn long-range dependencies.
Supervised by: Raghavendra Selvan