Hybrid Parallelization for Accelerating Visibility-Graph Construction and Community Detection on Temporal Data
Year: 2024
Type: book-chapter
Abstract: In the present era, vast amounts of time series data, particularly in biology, require efficient analysis due to high-dimensional datasets from advanced technologies. Transforming time series into networks and applying community detection methods can uncover dynamic patterns as temporal network communities. However, the large size of these datasets often extends analysis time. High Performance Computing (HPC) addresses this by accelerating traditional applications. This article presents an HPC-optimized version of the visibility-graph-based temporal community detection method. Enhancing the original algorithm with parallel processing, including shared memory and message passing models, improves adaptability. Experiments using an artificial sine series and two real-world datasets—biological time series and power consumption patterns—on Clemson's Palmetto Cluster demonstrate significant performance improvements and scalability over the original approach.
Source: Chronicle of computing
Authors: Xun Jia, Minzhang Zheng, Liwei Wang, Shuangshuang Jin
Institutions: Clemson University, St. Jude Children's Research Hospital
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4403290647