<ul><li>A new benchmark dataset called RelSC has been introduced for graph regression tasks.</li><li>RelSC is built from program graphs that combine syntactic and semantic information extracted from source code, and each graph is labeled with the execution-time cost of the program.</li><li>RelSC is released in two variants - RelSC-H with rich node features under a single edge type and RelSC-M that preserves the original multi-relational structure.</li><li>The dataset aims to provide a challenging and versatile benchmark for advancing graph regression methods by evaluating different graph neural network architectures on both variants of RelSC.</li></ul>

A Benchmark Dataset for Graph Regression with Homogeneous and Multi-Relational Variants

Discover more