This project concerns efficient parallel algorithms and software for emerging and future data analysis and mining applications, based on an emerging class of techniques known as tensor networks. Tensors, which are higher-dimensional generalizations of matrices, are finding applications in signal and image processing, computer vision, healthcare analytics, and neuroscience, to name just a few. Yet despite this demand, there is no comprehensive, high-performance software infrastructure targeting server systems that may have many parallel processors. Thus, the overarching research goal of this project is to design the first such infrastructure. The resulting prototype will be an open-source package, called the Parallel Tensor Infrastructure, or ParTI! The broader impact of the ParTI! project is to make the use of tensors, in a variety of data processing domains, much easier to do and more widespread.

The ParTI! project will focus specifically on algorithmic and software support for sparse tensors on single-node multi- and many-core accelerated platforms. The technical approach relies on a specific way of representing tensors, referred to as tensor networks. A tensor network is an efficient approach for representing the structure of a high-order tensor or tensor factorization. It can used by the data analyst as a simple, high-level way to express the specific structure or relationships he or she seeks in the data that the tensor represents. However, a tensor network is not just a tool for the analyst; it is also an abstract intermediate form, from which it is possible to derive algorithms, express and manage parallelism, and semi-automatically generate tensor processing software. This insight, combined with well-known data layout and communication-avoiding parallelization techniques from high-performance sparse linear algebra, is what will enable a ParTI! for tensor-based data analysis. The project will show the utility of this approach by evaluating the ParTI! prototype on real data sets and systems, through collaborations with government research laboratory and industry partners.

For further information see the project web site at: parti-project.org

Project Start
Project End
Budget Start
2015-09-01
Budget End
2021-08-31
Support Year
Fiscal Year
2015
Total Cost
$750,000
Indirect Cost
Name
Georgia Tech Research Corporation
Department
Type
DUNS #
City
Atlanta
State
GA
Country
United States
Zip Code
30332