Abstract. We obtain a Central Limit Theorem for a general class of additive parameters (costs, observables) associated to three standard Euclidean algorithms, with optimal speed of convergence. We also provide very precise asymptotic estimates and error terms for the mean and variance of such parameters. For costs that are lattice (including the number of steps), we go further and establish a Local Limit Theorem, with optimal speed of convergence. We view an algorithm as a dynamical system restricted to rational inputs, and combine tools imported from dynamics, such as transfer operators, with various other techniques: Dirichlet series, Perron’s formula, quasi-powers theorems, and the saddle-point method. Such dynamical analyses had previously been used to perform the average-case analysis of algorithms. For the present (dynamical) analysis in distribution, we require estimates on transfer operators when a parameter varies along vertical lines in the complex plane. To prove them, we adapt techniques introduced recently by Dolgopyat in the context of continuous-time dynamics [16]. 1.

Abstract. We study a general class of Euclidean algorithms which compute the greatest common divisor [gcd], and we perform probabilistic analyses of their main parameters. We view an algorithm as a dynamical system restricted to rational inputs, and combine tools imported from dynamics, such as transfer operators, with various tools of analytic combinatorics: generating functions, Dirichlet series, Tauberian theorems, Perron’s formula and quasi-powers theorems. Such dynamical analyses can be used to perform the average-case analysis of algorithms, but also (dynamical) analysis in distribution. 1. Introduction. Computing the Greatest Common Divisor [Gcd

In (V1), the author studies statistical properties of words generated by dynamical sources. This is done using generalized Ruelle operators. The aim of this article is to generalize the notion of sources for which the results hold. First, we avoid the use of Grothendieck theory and Fredholm determinants, this allows dynamical sources that cannot be extended to a complex disk or that are not analytic. Second, we consider Markov sources: the language generated by the source over an alphabet M is not necessarily M ∗.

This article is a continuation of our previous Algorithmic Column [54] (EATCS, 77, 2002) dedicated to activities of the Analysis of Algorithms group during the \Dagstuhl{ Period" (1993-1997). The rst three meetings took place in Schloss Dagstuhl, Germany.

The main justification for this book is that there have been significant advances in continued fractions over the past decade, but these remain for the most part scattered across the literature, and under the heading of topics from algebraic number theory to theoretical plasma physics. We now have a better understanding of the rate at which assorted continued fraction or greatest common denominator (gcd) algorithms complete their tasks. The number of steps required to complete a gcd calculation, for instance, has a Gaussian normal distribution. We know a lot more about badly approximable numbers. There are several related threads here. A badly approximable number is a number x such that {q|p − qx|: p, q ∈ Z and q ̸ = 0} is bounded below by a positive constant; badly approximable numbers have continued fraction expansions with bounded partial quotients, and so we are led to consider a kind of Cantor set EM consisting of all x ∈ [0, 1] such that the partial quotients of x are bounded above by M. The notion of a badly approximable rational number has the ring of crank mathematics, but it is quite natural to study the set of rationals r with partial quotients bounded by M. The number of such rationals with denominators up to n, say, turns out to be closely related to the Hausdorff dimension of EM, (comparable to n2dimEM) which is in turn related to the spectral radius of linear operators LM,s, acting on some suitably chosen space of functions f, and given by LM,sf(t) = ∑m k=1 (k + t) −sf(1/(k + t)). Similar operators have been studied by, among others, David Ruelle, in connection with theoretical one-dimensional plasmas, and they are related to entropy. Alongside these developments there has been a dramatic increase in the computational power available to investigators. This has been helpful on the theoretical side, as one is more likely to seek a proof for a result when,

www.imstat.org/aihp

Abstract. For large N, we consider the ordinary continued fraction of x = p/q with 1 ≤ p ≤ q ≤ N, or, equivalently, Euclid’s gcd algorithm for two integers 1 ≤ p ≤ q ≤ N, putting the uniform distribution on the set of p and qs. We study the distribution of the total cost of execution of the algorithm for an additive cost function c on the set Z ∗ + of possible digits, asymptotically for N → ∞. If c is nonlattice and satisfies mild growth conditions, the local limit theorem was proved previously by the second named author. Introducing diophantine conditions on the cost, we are able to control the speed of convergence in the local limit theorem. We use previous estimates of the first author and Vallée, and we adapt to our setting bounds of Dolgopyat and Melbourne on transfer operators. Our diophantine condition is generic. For smooth enough observables (depending on the diophantine condition) we attain the optimal speed.