English

SpectralUnmix: A Torch-Based Regularized Non-negative Matrix Factorization

Instrumentation and Methods for Astrophysics 2026-04-06 v1

Abstract

We present SpectralUnmix, an R package for regularized non-negative matrix factorization (NMF), implemented in torch with optional GPU acceleration. The package estimates low-rank non-negative representations through proximal-gradient updates and allows smoothness regularization along the spectral axis. As a compact demonstration, we apply the method to a subset of stellar spectra and compare the recovered NMF components with principal-component directions and representative stellar spectra. The package is released under the MIT license at \href{https://rafaelsdesouza.github.io/SpectralUnmix/}{this repository}.

Cite

@article{arxiv.2603.11111,
  title  = {SpectralUnmix: A Torch-Based Regularized Non-negative Matrix Factorization},
  author = {Rafael S. de Souza and Paula Coelho and Niranjana P and Ana L. Chies-Santos and Rogério Riffel},
  journal= {arXiv preprint arXiv:2603.11111},
  year   = {2026}
}

Comments

Accepted in Research Notes of AAS

R2 v1 2026-07-01T11:15:15.439Z