High-Performance Static and Streaming Tensor Factorization Algorithms
Abstract
The major goals of this project are the following: - Develop efficient and scalable serial and parallel algorithms for factorizing large, irregular, and sparse tensors. The research will address issues related to memory and operation-efficient computations, streaming data and computations, dynamic sparsity, and cache- and locality friendly tensor reorderings and partitionings. - Develop efficient and scalable serial and parallel algorithms for irregular computations arising when operating on large, sparse, and unstructured static and dynamic graphs. Examples of such computations arise in various network-science-related fundamental operations (e.g., triangle counting, truss-decomposition, clustering) and in computations arising in applications involving graph signal processing and graph neural network.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 04, 2023
- Accession Number
- AD1210713
Entities
People
- George Karypis
Organizations
- University of Minnesota