SVD Factorization for Tall-and-Fat Matrices on Parallel Architectures
We demonstrate an implementation for an approximate rank-k SVD factorization, combining well-known randomized projection techniques with previously known paralel solutions in order to compute steps of the random projection based SVD procedure. We structure the problem in a way that it reduces to fast computation around $k \times k$ matrices computed on a single machine, greatly easing the computability of the problem. The paper is also a tutorial on paralel linear algebra methods using a plain architecture without burdensome frameworks.
PDF AbstractCategories
Distributed, Parallel, and Cluster Computing
Datasets
Add Datasets
introduced or used in this paper