The paper addresses the target stationary distribution problem (TSDP), which aims to find a minimum norm perturbation of an irreducible stochastic matrix G such that the perturbed matrix has a prescribed target stationary distribution.
The key contributions are:
For the case where the support of the perturbation is constrained to the non-zero entries of G plus the diagonal, the paper provides a closed-form feasible solution that minimizes the component-wise ℓ1 norm. This solution can be rank-1 under certain conditions on the target distribution.
For the general case with an arbitrary support constraint Ω, the paper proposes an efficient linear optimization formulation of the TSDP that has only 2n equality constraints and less than 2|Ω| bounded variables. This allows solving large-scale sparse problems efficiently.
The paper analyzes the properties of the solutions, including their sparsity, optimality, and the impact of reordering the stationary distribution vector μ to better match the target distribution μ̂.
Numerical experiments demonstrate the effectiveness of the proposed approach, showing the ability to solve sparse problems of size up to 105 × 105 in a few minutes.
To Another Language
from source content
arxiv.org
Principais Insights Extraídos De
by Nicolas Gill... às arxiv.org 04-29-2024
https://arxiv.org/pdf/2312.16011.pdfPerguntas Mais Profundas