Distributed-Memory Parallel Symmetric Nonnegative Matrix Factorization

被引：3

作者：

Eswar, Srinivas ^{[1
]}

Hayashi, Koby ^{[1
]}

Ballard, Grey ^{[2
]}

Kannan, Ramakrishnan ^{[3
]}

Vuduc, Richard ^{[1
]}

Park, Haesun ^{[1
]}

机构：

[1] Georgia Inst Technol, Dept Computat Sci & Engn, Atlanta, GA 30332 USA

[2] Wake Forest Univ, Dept Comp Sci, Winston Salem, NC 27101 USA

[3] Oak Ridge Natl Lab, Computat Data Analyt Grp, Oak Ridge, TN USA

来源：

PROCEEDINGS OF SC20: THE INTERNATIONAL CONFERENCE FOR HIGH PERFORMANCE COMPUTING, NETWORKING, STORAGE AND ANALYSIS (SC20) | 2020年

关键词：

High performance computing; Newton method; Parallel algorithms; Symmetric Matrices; COLLECTIVE COMMUNICATION; COORDINATE DESCENT; ALGORITHMS;

D O I：

10.1109/sc41405.2020.00078

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

We develop the first distributed -memory parallel implementation of Symmetric Nonnegative Matrix Factorization (SymNMF), a key data analytics kernel 14 clustering and dimensionality reduction. Our implementation includes two different algorithms for SytnNMF, which give comparable results in terms of time and accuracy. The first algorithm is a parallelization of an existing sequential approach that uses solvers for nonsymmetric NNW The second algorithm is a novel approach based on the Gauss -Newton method. It exploits second -order information without incurring large computational and memory costs. We evaluate the scalability of our algorithms on the Summit system at Oak Ridge National Laboratory, scaling up to 128 nodes (4,096 cores) with 70% efficiency. Additionally, we demonstrate our software on an image segmentation task.

引用

页数：14