English

Assigning Stationary Distributions to Sparse Stochastic Matrices

Numerical Analysis 2025-01-10 v3 Numerical Analysis Optimization and Control Probability Computation

Abstract

The target stationary distribution problem (TSDP) is the following: given an irreducible stochastic matrix GG and a target stationary distribution μ^\hat \mu, construct a minimum norm perturbation, Δ\Delta, such that G^=G+Δ\hat G = G+\Delta is also stochastic and has the prescribed target stationary distribution, μ^\hat \mu. In this paper, we revisit the TSDP under a constraint on the support of Δ\Delta, that is, on the set of non-zero entries of Δ\Delta. This is particularly meaningful in practice since one cannot typically modify all entries of GG. We first show how to construct a feasible solution G^\hat G that has essentially the same support as the matrix GG. Then we show how to compute globally optimal and sparse solutions using the component-wise 1\ell_1 norm and linear optimization. We propose an efficient implementation that relies on a column-generation approach which allows us to solve sparse problems of size up to 105×10510^5 \times 10^5 in a few minutes. We illustrate the proposed algorithms with several numerical experiments.

Keywords

Cite

@article{arxiv.2312.16011,
  title  = {Assigning Stationary Distributions to Sparse Stochastic Matrices},
  author = {Nicolas Gillis and Paul Van Dooren},
  journal= {arXiv preprint arXiv:2312.16011},
  year   = {2025}
}

Comments

29 pages, code available from https://gitlab.com/ngillis/TSDP. In this third version, we have added clarifications, corrections and remarks suggested to us by anonymous reviewers

R2 v1 2026-06-28T14:02:03.803Z