An end-to-end bi-objective approach to deep graph partitioning

Pengcheng Wei, Yuan Fang, Zhihao Wen, Zheng Xiao, Binbin Chen

Abstract: Highlights•The first GNN-based end-to-end bi-objective graph partitioning method.•Exploit multilevel (node-, local, global) graph features.•Propose the partition-wise balance function.•Propose the Hardmax operator for end-to-end optimization.•Test our method against 9 baselines on 11 datasets.