Parallel Algorithms for Computing and Applications of the Dense Canonical Polyadic Decomposition
Author | : Koby B. Hayashi |
Publisher | : |
Total Pages | : 80 |
Release | : 2018 |
ISBN-10 | : OCLC:1044768431 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book Parallel Algorithms for Computing and Applications of the Dense Canonical Polyadic Decomposition written by Koby B. Hayashi and published by . This book was released on 2018 with total page 80 pages. Available in PDF, EPUB and Kindle. Book excerpt: Tensor decompositions have gained popularity in various research communities as a means of analyzing high dimensional, complex data. Generalizations of matrix decompositions to more than 2 dimensions, tensor decompositions are useful in machine learning, chemometrics, computer vision, graph analysis, and other areas/applications. The focus of this Masters Thesis is the Canonical Polyadic or CP Decomposition for dense tensors. We discuss and present shared and distributed memory parallel algorithms for computing a CP Decomposition and explore a motivating neuroimaging application. Our implementations scale to hundreds of nodes and thousands of cores on distributed memory machines. We obtain up to 2× speed up over state of the art software, and demonstrate the CP’s utility in revealing latent pattern in data via a neuroimaging application.