Support data for the paper "Addressing Large Rank-Deficient Linear Least-Squares Problems on Shared-Memory CPU and GPU Architectures"

These files correspond to the supporting data used in Section 4.1 of the paper "Addressing Large Rank-Deficient Linear Least-Squares Problems on Shared-Memory CPU and GPU Architectures".We provide three test cases generated by the authors: 1. Toeplitz.mat A Toeplitz matrix constructed to study a case of approximately low rank. Format: MATLAB sparse matrix Dimensions: m=15,000, n=14,000 Numerical rank: 14,000 This matrix exhibits rapidly decaying singular values and is intended to represent a structured, approximately low-rank scenario. 2. Tomography1.mat A system matrix (stored as sparse to save space) and right-hand side vector b corresponding to a 2D Computed Tomography (CT) forward model. Matrix format: MATLAB sparse matrix Dimensions: m=148,672, n=147,456 Numerical rank: 141,270 b: projection vector (sinogram) for one slice of the Forbild Head Phantom The data corresponds to a simulated CT scan with: 736 detectors 202 projections Reconstruction grid of 384×384 pixels Forward model generated using Joseph’s method 3. Tomography2.mat A larger tomography test case. Matrix format: MATLAB sparse matrix Dimensions: m=590,272, n=589,824 Numerical rank: 560,526 b: sinogram for one slice of the Forbild Head Phantom Associated with: 736 detectors 802 projections Reconstruction grid of 768×768 Forward model generated using Joseph’s method