The quality of the diophantine approximations found by the Jacobi--Perron algorithm and related algorithms (Q688961)

The author introduces two classes of algorithms which extend the multivariate continued fraction approximation. The two algorithms are called Markovian and linear simplex splitting multivariate continued fraction algorithms, respectively, and they are built on previous concepts of multivariate continued fractions. In these new algorithms the memoryless property of the \(n\)th iterative of the algorithm is stressed (hence, Markovian), and the term `simplex splitting' refers to the structure of the algorithm: a cone is split into integral subcones and the approximating one is selected in the algorithm. The author uses Oseledec's multiplicative ergodic theorem to produce a representation of both the uniform approximation exponent and the best approximation exponent both of which are shown to be constant on a set of Lebesgue measure one.