Abstraction-guided truncations for stationary distributions of Markov population models (Q832087)

State-of-the-art methods for numerically calculating the stationary distribution of Markov population models rely on coarse truncations of irrelevant parts of large or infinite discrete state-spaces. These truncations are either obtained from the stationary statistical moments of the process or from Lyapunov theory. They are limited in shape because these methods do not take into account the detailed steady-state flow within the truncated state-space but only consider the average drift or stationary moments. In this article, the authors propose a method to find a tight truncation that is not limited in its shape and iteratively optimizes the set based on numerically cheap solutions of abstract intermediate models. It captures the main portion of probability mass even in the case of complex behaviors efficiently. In particular, the method represents another option, where Lyapunov analysis leads to forbiddingly large truncations. An aggregation scheme similar to the one used here has been previously proposed in [\textit{M. Backenköhler} et al., Lect. Notes Comput. Sci. 12651, 210--229 (2021; Zbl 1472.60140)] to analyze the bridging problem on Markov population models. This is the problem of analyzing process dynamics under both initial and terminal constraints. For the entire collection see [Zbl 1482.68004].