An adaptive aggregation/disaggregation algorithm for hierarchical Markovian models (Q1610177): Difference between revisions

The analysis of continuous time Markov chains with huge state spaces requires considerable numerical effort. The author develops an algorithm for the evaluation of the stationary probabilities vector applicable to a quite general class of models. Hierarchically structured chains can be described by a generator matrix with a compact tensor representation. The algorithm combines SOR iterations exploiting the tensor structure with adaptive aggregation/disaggregation steps. Several examples are given to compare the performance of different variants of the algorithm.