A paintbox characterization of labeled partitions
Abstract
We study pairs of labeled random partitions (Π,τ) with a finite label set E, which can be thought of as a random mutation and coalescence mechanism of a countable number of blocks. Labeled exchangeable partition probability functions (LEPPFs) are defined via labeled analogues of Kingman's paintbox partitions. They satisfy normalization, symmetry and consistency in a multivariate sense. Under an equivariance condition, a representation for (Π,τ) in terms of labeled paintbox partitions is derived, leading to a full characterization of LEPPFs by a unique probability measure G. This representation is extended to jointly exchangeable and consistent transition pairs using a product measure F ⊗ G. These results yield a one-to-one correspondence between distributions of jointly exchangeable and consistent transition pairs and pairs of exchangeable distributions and LEPPFs.
Related
BibTeX
@misc{flamm2026paintbox,
title = {{A paintbox characterization of labeled partitions}},
author = {Maximilian Flamm and Martin Möhle},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.29824},
eprint = {2610.29824},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.29824},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A paintbox characterization of labeled partitions
AU - Maximilian Flamm
AU - Martin Möhle
PY - 2026
DA - 2026/10/22
PB - arXiv
DO - 10.48550/arxiv.2610.29824
UR - https://arxiv.org/abs/2610.29824
AB - We study pairs of labeled random partitions (Π,τ) with a finite label set E, which can be thought of as a random mutation and coalescence mechanism of a countable number of blocks. Labeled exchangeable partition probability functions (LEPPFs) are defined via labeled analogues of Kingman's paintbox partitions. They satisfy normalization, symmetry and consistency in a multivariate sense. Under an equivariance condition, a representation for (Π,τ) in terms of labeled paintbox partitions is derived, leading to a full characterization of LEPPFs by a unique probability measure G. This representation is extended to jointly exchangeable and consistent transition pairs using a product measure F ⊗ G. These results yield a one-to-one correspondence between distributions of jointly exchangeable and consistent transition pairs and pairs of exchangeable distributions and LEPPFs.
ER -