(eng): In the article algorithm for splitting and merging complexes of convex polyhedra is proposed. The need for such an algorithm emerges for example in computation of normalization factor of Bayesian estimated posterior likelihood of parameters in an auto-regression model with Laplace distributed innovations, where the split of the complex corresponds to newly arrived data incorporation and the merging relates to making such a model adaptive by forgetting older data values -- the situation is most apparent in case of an estimation on a moving window.