In this paper two bottom-up optimization algorithms suitable for growing classification and regression trees are outlined and their properties studied.