The classical continued fraction algorithm as well as its generalizations to higher dimensions (like, for instance, Brun's algorithm or the Jacobi-Perron algorithm) produce sequences of unimodular matrices. Our aim is to study such sequences and to discuss convergence properties of them. These convergence properties are formulated in terms of hyperbolicity properties. We want to associate Markov partitions to these continued fraction algorithms. Indeed, it has been known for a long time that we can associate a Markov partition to a single hyperbolic matrix. It turns out that it is also possible to associate Markov partitions with a "hyperbolic" sequence of matrices, and, hence, with a strongly convergent continued fraction algorithm. To define such "nonstationary" Markov partitions, we relate a sequence of substitutions to a sequence of matrices. This will lead to a symbolic coding of such a sequence. This is joint work with P. Arnoux, V. Berthé, M. Minervino, and W. Steiner.