software . 2021

Out-of-memory SVD

Ghennupati, Gopinath; Carrillo Cabada, Hector; Skau, Erik; Alexandrov, Boian; Djidjev, Hristo;
Restricted
  • Published: 17 Jun 2021
  • Publisher: DOE CODE
Abstract
Singular value decomposition (SVD) is a matrix factorization widely used for dimensionality reduction, data analytics, information retrieval and unsupervised learning. In the SVDs applications for big-data, usually, only the singular values are calculated. However, new methods, such as the tensor network factorization, require an accurate retrieval of a substantial number of singular vectors for truncated SVD. Also, many real-world datasets are too big to fit directly into the memory, which mandates the development of out-of-memory algorithms that work with data that is primarily on the disk. Here, building upon a previous work, we present a method for computati...
Subjects
ACM Computing Classification System: MathematicsofComputing_NUMERICALANALYSIS
Any information missing or wrong?Report an Issue