A parallel ADMM-based convex clustering method

Abstract Convex clustering has received recently an increased interest as a valuable method for unsupervised learning. Unlike conventional methods such k-means, its formulation corresponds to solving convex optimization problem and hence, alleviates initialization local minima problems. However, while several algorithms have been proposed solve formulations, including those based on the alternating direction of multipliers (ADMM), there is currently limited body work developing scalable parallel distributed solvers clustering. In this paper, we develop parallel, ADMM-based method, modified sum-of-norms (SON) master–worker architectures, where data be clustered are partitioned across number worker nodes, provide efficient, open-source implementation (available Parallel https://github.com/lidijaf/Parallel-ADMM-based-convex-clustering . Accessed 10 June 2022) high-performance computing (HPC) cluster environments. Extensive numerical evaluations real synthetic sets demonstrate high degree scalability efficiency when compared with existing alternative