On two proofs of d2 mixing of weighted Dikin walks
Abstract
We study the mixing time of weighted Dikin walks for sampling from exponential distributions on polytopes and truncated positive-semidefinite (PSD) cones. Our first result gives a general total-variation mixing bound under strong self-concordance, ν̅-symmetry, and mixed-trace regularity on the local metric. The key idea is to control the Metropolis--Hastings acceptance probability on a high-probability region rather than at every point. Applying this framework to the Lee--Sidford, Lewis-weight, and John metrics yields an Õ(d2) mixing bound for sampling from polytopes, while applying it to a hybrid barrier yields an Õ(d4) mixing bound for sampling from truncated PSD cones. Our second result establishes stronger χ2-divergence guarantees and pointwise acceptance control using a new fourth-order bootstrap condition. For a suitably scaled Lee--Sidford metric, this yields an Õ(d2) mixing bound in χ2-divergence, improving on the previous Õ(d9/4) bound.
Related
BibTeX
@misc{chen2026proofs,
title = {{On two proofs of $d^2$ mixing of weighted Dikin walks}},
author = {Yuansi Chen and Yunbum Kook},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.28566},
eprint = {2608.28566},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.28566},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - On two proofs of d^2 mixing of weighted Dikin walks
AU - Yuansi Chen
AU - Yunbum Kook
PY - 2026
DA - 2026/08/28
PB - arXiv
DO - 10.48550/arxiv.2608.28566
UR - https://arxiv.org/abs/2608.28566
AB - We study the mixing time of weighted Dikin walks for sampling from exponential distributions on polytopes and truncated positive-semidefinite (PSD) cones. Our first result gives a general total-variation mixing bound under strong self-concordance, ν̅-symmetry, and mixed-trace regularity on the local metric. The key idea is to control the Metropolis--Hastings acceptance probability on a high-probability region rather than at every point. Applying this framework to the Lee--Sidford, Lewis-weight, and John metrics yields an Õ(d^2) mixing bound for sampling from polytopes, while applying it to a hybrid barrier yields an Õ(d^4) mixing bound for sampling from truncated PSD cones. Our second result establishes stronger χ^2-divergence guarantees and pointwise acceptance control using a new fourth-order bootstrap condition. For a suitably scaled Lee--Sidford metric, this yields an Õ(d^2) mixing bound in χ^2-divergence, improving on the previous Õ(d^9/4) bound.
ER -