The Eigenvalues Entropy as a Classifier Evaluation Measure

Doulaye Dembélé · 2025

Read The Eigenvalues Entropy as a Classifier Evaluation Measure by Doulaye Dembélé (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01904.txt.

Description

Doulaye Dembélé proposes judging classifiers by the entropy of their eigenvalues. Classification quality is usually scored from a contingency table, but many standard measures misbehave when the classes are imbalanced, exactly the situation in face recognition, text mining, and medical pattern tasks. Dembélé's measure builds the score from the eigenvalues of the problem's natural matrix summary, summarized through their entropy.

For binary problems, the paper derives relations between the eigenvalues and familiar quantities: sensitivity, specificity, the area under the receiver operating curve, and the Gini index. A side result estimates the confusion matrix in a way meant to blunt the curse of imbalanced classes. Worked data examples are used to argue that the eigenvalue entropy measure outperforms the gold-standard measures it is compared against, particularly when class proportions are skewed.

Catalog record

CREATORDoulaye Dembélé
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-01904

Browse the full library · About the Tower of Babel