“…However, here we propose a different argument, which allows us to consider the case of nonlinear permutations, and in fact we also obtain an improvement of the result of [7] as well. These estimates for almost all initial values are much stronger and hold in a much wider range than those obtained for every initial value; see [6] and [10] for the case of inversive and polynomial generators, respectively. We note that besides being of theoretical interest, such results that apply to almost all initial values are probably most relevant for practical applications of pseudorandom generators, in both cryptography and numerical analysis.…”