English

Structured Variational $D$-Decomposition for Accurate and Stable Low-Rank Approximation

Numerical Analysis 2025-06-11 v1 Machine Learning Numerical Analysis

Abstract

We introduce the DD-decomposition, a non-orthogonal matrix factorization of the form APDQA \approx P D Q, where PRn×kP \in \mathbb{R}^{n \times k}, DRk×kD \in \mathbb{R}^{k \times k}, and QRk×nQ \in \mathbb{R}^{k \times n}. The decomposition is defined variationally by minimizing a regularized Frobenius loss, allowing control over rank, sparsity, and conditioning. Unlike algebraic factorizations such as LU or SVD, it is computed by alternating minimization. We establish existence and perturbation stability of the solution and show that each update has complexity O(n2k)\mathcal{O}(n^2k). Benchmarks against truncated SVD, CUR, and nonnegative matrix factorization show improved reconstruction accuracy on MovieLens, MNIST, Olivetti Faces, and gene expression matrices, particularly under sparsity and noise.

Keywords

Cite

@article{arxiv.2506.08535,
  title  = {Structured Variational $D$-Decomposition for Accurate and Stable Low-Rank Approximation},
  author = {Ronald Katende},
  journal= {arXiv preprint arXiv:2506.08535},
  year   = {2025}
}
R2 v1 2026-07-01T03:08:36.687Z