Towards information-theoretic visualization evaluation measure: a practical example for Bertin's matrices

Abstract

This paper presents a discussion about matrix-based representation evaluation measures, including a review of related evaluation measures from different scientific disciplines and a proposal for promising approaches. The paper advocates linking or replacing a large portion of indefinable aesthetics with a mathematical framework and theory backed up by an incomputable function -- Kolmogorov complexity. A suitable information-theoretic evaluation measure is proposed together with a practical approximating implementation example for Bertin's Matrices.