Exact entropy constants for balanced multislices
Consider the uniform measure $\mu$ on words with $m$ colours, each appearing $r$ times, and assign unit weight to every unordered pair of coordinates. If $Q(f)$ is the sum of the pair conditional entropies, we prove that the largest coefficient $\kappa$ in $\kappa\operatorname{Ent}_\mu(f)\leq Q(f)$ is $\binom m2r^2\log 2/\log((mr)!/(r!)^m)$. A singleton indicator attains equality. This establishes the balanced case of the multicolour formula proposed by Bristiel and Caputo and determines the optimal logarithmic edge-expansion coefficient of the associated transposition graph. The proof combines their exact Bernoulli--Laplace entropy constant with Salez's conditioning on the complete position set of one colour. Equal multiplicities make the resulting induction close without loss. We also record the resulting classical log-Sobolev bounds and give a self-contained proof of the known binary input.
Authors
- Sam Vaseghi (ORCID: https://orcid.org/0009-0006-4105-1949)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24247348
- Primary Topic
- Probability and Statistical Research
- Type
- preprint