MetaSieve: Faster Relational Deep Learning through SQL-Based Metapath Selection
What it is
MetaSieve is a method for optimizing Relational Deep Learning (RDL) by pruning uninformative metapaths in graph neural network training. It uses SQL-based statistics and a scoring function to select lightweight, informative metapaths, reducing subgraph size and training time without sacrificing accuracy.
The problem it solves
High training costs in Relational Deep Learning caused by the large size of sampled subgraphs during GNN training.
What you could use it for
- Accelerating GNN training on large relational databases
- Optimizing subgraph sampling for RelBench benchmarks
- Integrating metapath selection into diverse GNN architectures for classification and regression
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.