Workshop: IA^3 2026 — 16th Workshop on Irregular Applications: Architectures and Algorithms
Authors: Arindam Khanda (Kennesaw State University), S. M. Shovan (Missouri University of Science and Technology), Sanjukta Bhowmick (University of North Texas), and Sajal K. Das (Missouri University of Science and Technology)
Abstract: Finding shortest paths is a well-studied graph problem, with parallel approaches for both static and dynamic settings. However, most existing methods target dyadic networks, where each edge represents an interaction between only two vertices. Many real-world systems involve group interactions, which are modeled more naturally by hypergraphs, where a hyperedge may connect any number of vertices. Moreover, these interactions are often time-dependent, leading to frequent topological changes in the underlying hypergraphs. Yet no scalable approach exists to efficiently update shortest paths, or hyperpaths, in large dynamic hypergraphs. To address this gap, we propose a parallel shortest hyperpath update method that uses a dynamic data structure to maintain the hypergraph, identifies the subgraphs affected by temporal changes, and efficiently updates previously computed shortest hyperpaths. Experiments show that our GPU-based implementation outperforms the recomputation-based baseline by an average speedup of 28.6x and up to 178.3x on large real-world hypergraphs.
Back to IA^3 2026 — 16th Workshop on Irregular Applications: Architectures and Algorithms Archive Listing
Back to Full Workshop Archive Listing