A High-Performance Singular Value Decomposition Solver for Large-Scale Problems in Machine Learning Kernel Methods

This is a collaborative research effort with computer science and machine learning researchers at the College of William and Mary. Recently, Machine Learning (ML) has emerged as one of the primary clients for SVD calculations. With the unprecedented growth in the size of data and models that need to be analyzed, the linear algebra tools for SVD that are based on dense matrix transformations and are ubiquitous in ML packages are now facing severe scalability problems. This research effort utilizes the XSEDE Stampede 2 supercomputer as a development and testing platform to develop scalable, parallel software for SVD problems.

Researchers at College of William and Mary

Oct 2019

Ongoing