English

On Maximum Entropy Linear Feature Inversion

Machine Learning 2024-07-22 v1

Abstract

We revisit the classical problem of inverting dimension-reducing linear mappings using the maximum entropy (MaxEnt) criterion. In the literature, solutions are problem-dependent, inconsistent, and use different entropy measures. We propose a new unified approach that not only specializes to the existing approaches, but offers solutions to new cases, such as when data values are constrained to [0, 1], which has new applications in machine learning.

Keywords

Cite

@article{arxiv.2407.14166,
  title  = {On Maximum Entropy Linear Feature Inversion},
  author = {Paul M Baggenstoss},
  journal= {arXiv preprint arXiv:2407.14166},
  year   = {2024}
}

Comments

Submitted IEEE-Signal Processing Letters