D-SLR: The Disjoint Row-Sparse plus Low-Rank Decomposition
D-SLR proposes a method to improve or match truncated SVD by either storing rows verbatim or approximating them with a low-rank fit, never both. It solves the difficult problem of jointly optimizing row sparsity and low-rank approximation. The method provides a closed-form algorithm that scores all possibilities without iteration, offering a certificate to guide selection.