An approach to the analysis of adaptation-algorithm dynamics is considered. It consists of constructing approximate models of the algorithms. A formal definition of the adaptive system is given. The models described are employed to investigate the dynamics of a no-search gradient algorithm, as well as a random-search algorithm. Estimates are obtained for the practical rate of convergence of these algorithms in the problem of adaptive identification of a static object.