"Scaling Subgraph Isomorphism: Exploiting Massive GPU Parallelism with VF-GPU"

"This dataset accompanies the paper \"Scaling Subgraph Isomorphism: Exploiting Massive GPU Parallelism with VF-GPU\" and provides the target and query graphs used to benchmark the proposed algorithm against state-of-the-art CPU- and GPU-based subgraph isomorphism methods (VF3, VF3-L, GSI, E-GSM). Three large-scale target graphs from the SNAP repository are included: DBLP (317,080 nodes, 1,049,866 edges, co-authorship network), Web-Google (875,713 nodes, 5,105,039 edges, hyperlink graph), and Live-Journal (4,847,571 nodes, 68,993,773 edges, social network). For each target graph, query graphs of size 8, 16, 32, and 64 nodes were generated via random-walk sampling, with smaller queries derived by node\/edge removal from the largest one. Each query size is paired with seven label-distribution configurations (64, 32, 16, 8, 4, 2 distinct labels, plus an original-labels setting), yielding 100 distinct queries per size\/label combination and a total of 2,800 queries per target graph (8,400 queries overall). The dataset is intended to support reproducibility of the VF-GPU evaluation and to serve as a benchmark for future subgraph isomorphism algorithms on large, real-world graphs."