A reversible automaton is a finite automaton in which eachletter induces a partial one-to-one map from the set of states intoitself. We solve the following problem proposed by Pin. Given an alphabet A, does there exist a sequence of languages K n on A which can be accepted by a reversible automaton, and such that the number of states of the minimal automaton of K n is in O(n), while the minimal number of states of a reversible automaton accepting K n is in O(ρ n ) for some ρ > 1? We give such an example with $\rho=\left(\frac{9}{8}\right)^{\frac{1}{12}}$ .

In this paper, we establish the following results: (i) every strongly finite dilator is finitely presentable in the category of endofunctors on the category of ordinals; (ii) a dilator F is strongly finite if and only if F is finitely presentable in the category of dilators.

The standard procedure to transform a regular expression of size n to an ϵ-free nondeterministic finite automaton yields automata with O(n) states and O(n2) transitions. For a long time this was supposed to be also the lower bound, but a result by Hromkovic et al. showed how to build an ϵ-free NFA with only O(n log2(n)) transitions. The current lower bound on the number of transitions is Ω(n log(n)). A rough running time estimation for the common follow sets (CFS) construction proposed by Hromkovič et al. yields a cubic algorithm. In this paper we present a sequential algorithm for the CFS construction which works in time O(n log(n) + size of the output). On a CREW PRAM the CFS construction can be performed in time O(log(n)) using O(n + (size of the output)/log(n)) processors. We also present a simpler proof of the lower bound on the number of transitions.

It is known that finding a perfect matching in a general graph is AC0-equivalent to finding a perfect matching in a 3-regular (i.e. cubic) graph. In this paper we extend this result to both, planar and bipartite cases. In particular we prove that the construction problem for perfect matchings in planar graphs is as difficult as in the case of planar cubic graphs like it is known to be the case for the famous Map Four-Coloring problem. Moreover we prove that the existence and construction problems for perfect matchings in bipartite graphs are as difficult as the existence and construction problems for a weighted perfect matching of O(m) weight in a cubic bipartite graph.

In this paper, we examine the class of "deco" polyominoes and the succession rule describing their construction. These polyominoes are enumerated according to their directed height by factorial numbers. By changing some aspects of the "factorial" rule, we obtain some succession rules that describe various "deco" polyomino subclasses. By enumerating the subclasses according to their height and width, we find the following well-known numbers: Stirling numbers of the first and second kind, Narayana and odd index Fibonacci numbers. We wish to point out how the changes made on the original succession rule yield some new succession rules that produce transcendental, algebraic and rational generating functions.

We prove that the Christofides algorithm gives a $\frac{4}{3}$ approximation ratio for the special case of traveling salesman problem (TSP) in which the maximum weight in the given graph is at most twice the minimum weight for the odd degree restricted graphs. A graph is odd degree restricted if the number of odd degree vertices in any minimum spanning tree of the given graph is less than $\frac{1}{4}$ times the number of vertices in the graph. We prove that the Christofides algorithm is more efficient (in terms of runtime) than the previous existing algorithms for this special case of the traveling salesman problem. Secondly, we apply the concept of stability of approximation to this special case of traveling salesman problem in order to partition the set of all instances of TSP into an infinite spectrum of classes according to their approximability.

Considering each occurrence of a word w in a recurrentinfinite word, we define the setof return words of w to be the set of all distinct words beginningwith an occurrence of wand ending exactly just before the next occurrence of w in the infiniteword. We give a simpler proof of therecent result (of the second author) that an infinite word is Sturmianif and only if each of its factors has exactly two return words in it.Then, considering episturmian infinite words, which are a naturalgeneralization of Sturmian words,we study the position of the occurrences of any factorin such infinite wordsand we determinate the return words. At last, we apply these results inorder to get a kind of balance property ofepisturmian words and to calculate the recurrence function of thesewords.

We proceed our work on iterated transductions by studying the closure underunion and composition of some classes of iterated functions. Weanalyze this closure for the classes of length-preservingrational functions, length-preserving subsequential functions andlength-preserving sequential functions with terminal states. All the classes weobtain are equal. We also study the connection with deterministiccontext-sensitive languages.

We introduce a domain-free λµ-calculus of call-by-value as a short-hand for the second order Church-style. Our motivation comes from the observation that in Curry-style polymorphic calculi, control operators such as callcc-operators cannot, in general, handle correctly the terms placed on the control operator's left, so that the Curry-style system can fail to prove the subject reduction property. Following the continuation semantics, we also discuss the notion of values in classical system, and propose an extended form of values. It is proved that the CPS-translation is sound with respect to domain-free λ2 (second-order λ-calculus). As a by-product, we obtain the strong normalization property for the second-order λµ-calculus of call-by-value in domain-free style. We also study the problems of type inference, typability, and type checking for the call-by-value system. Finally, we give a brief comparison with standard ML plus callcc, and discuss a natural way to avoid the unsoundness of ML with callcc.

We show that the FIX type theory introduced by Crole and Pitts [3] can be encoded in variants of Abadi and Cardelli's object calculi. More precisely, we show that the FIX type theory presented with judgements of both equality and operational reduction can be translated into object calculi, and the translation proved sound. The translations we give can be seen as using object calculi as a metalanguge within which FIX can be represented; an analogy can be drawn with Martin Löf's Theory of Arities and Expressions. As well as providing a description of certain interesting recursive objects in terms of rather simpler expressions found in the FIX type theory, the translations will be of interest to those involved with the automation of operational semantics.

In this paper we investigate the average Horton-Strahler number of all possible tree-structures of binary tries. For that purpose we consider a generalization of extended binary trees where leaves are distinguished in order to represent the location of keys within a corresponding trie. Assuming a uniform distribution for those trees we prove that the expected Horton-Strahler number of a tree with α internal nodes and β leaves that correspond to a key is asymptotically given by $$\frac{4^{2\beta-\alpha}\log(\alpha)(2\beta-1)(\alpha+1)(\alpha+2){2\alpha+1\choose \alpha-1}}{8\sqrt{\pi}\alpha^{3/2}\log(2)(\beta-1)\beta{2\beta\choose \beta}^2}$$ provided that α and β grow in some fixed proportion ρ when α → ∞ . A similar result is shown for trees with α internal nodes but with an arbitrary number of keys.

The purpose of this paper is to show connections between iterated length-preserving rational transductions and linear space computations. Hence, we study the smallest family of transductions containing length-preserving rational transductions and closed under union, composition and iteration. We give several characterizations of this class using restricted classes of length-preserving rational transductions, by showing the connections with "context-sensitive transductions" and transductions associated with recognizable picture languages.

We consider languages expressed by word equations in two variables and give a complete characterization for their complexity functions, that is, the functions that give the number of words of the same length. Specifically, we prove that there are only five types of complexities: constant, linear, exponential, and two in between constant and linear. For the latter two, we give precise characterizations in terms of the number of solutions of Diophantine equations of certain types. In particular, we show that the linear upper bound on the non-exponential complexities by Karhumäki et al. in [9], is tight. There are several consequences of our study. First, we derive that both of the sets of all finite Sturmian words and of all finite Standard words are expressible by word equations. Second, we characterize the languages of non-exponential complexity which are expressible by two-variable word equations as finite unions of several simple parametric formulae and solutions of a two-variable word equation with a finite graph. Third, we find optimal upper bounds on the solutions of (solvable) two-variable word equations, namely, linear bound for one variable and quadratric for the other. From this, we obtain an $\mathcal{O}(n^6)$ algorithm for testing the solvability of two-variable word equations, improving thus very much Charatonik and Pacholski's $\mathcal{O}(n^100)$ algorithm from [3].

The complexity of computing, via threshold circuits, the iterated product and powering of fixed-dimension $k\times k$ matrices with integer or rational entries is studied. We call these two problems $\sf IMP_k$ and $\sf MPOW_k$, respectively, for short. We prove that: (i) For $k\geq2$, $\sf IMP_k$ does not belong to ${\rm TC}^0$, unless ${\rm TC}^0={\rm NC}^1$.newline (ii) For stochastic matrices : $\sf IMP_2$ belongs to ${\rm TC}^0$ while, for $k\geq3$, $\sf IMP_k$ does not belong to ${\rm TC}^0$, unless ${\rm TC}^0={\rm NC}^1$. (iii) For any k, $\sf MPOW_k$ belongs to ${\rm TC}^0$.

We construct, for each integer n, three functions from {0,1}n to {0,1} such that any boolean mapping from {0,1}n to {0,1}n can be computed with a finite sequence of assignations only using the n input variables and those three functions.

This paper deals with lower bounds on the approximability of different subproblems of the TravelingSalesman Problem (TSP) which is known not to admit any polynomial time approximation algorithm in general(unless $\mathcal{P}=\mathcal{NP}$ ). First of all, we present an improved lower bound for the Traveling Salesman Problemwith Triangle Inequality, Delta-TSP for short. Moreover our technique, an extension of the method ofEngebretsen [11], also applies to the case of relaxed and sharpened triangle inequality,respectively, denoted $\Delta_\beta$ -TSP for an appropriate β. In case of theDelta-TSP, we obtain a lowerbound of $\frac{3813}{3812}-\varepsilon$ on the polynomial-time approximability (for any small $\varepsilon> 0$ ), compared to the previous bound of $\frac{5381}{5380}-\varepsilon$ in [11]. Incase of the $\Delta_\beta$ -TSP, for the relaxed case ( $\beta> 1$ ) we present a lower bound of $\frac{3803+10\beta}{3804+8\beta}-\varepsilon$ , and for the sharpened triangle inequality( $\frac{1}{2}< \beta< 1$ ), the lower bound is $\frac{7611+10\beta^2+5\beta}{7612+8\beta^2+4\beta}-\varepsilon$ . The latter result is of interestespecially since it shows that the TSP is $\mathcal{APX}$ -hard even if one comes arbitrarily close to the trivialcase that all edges have the same cost.

We show that one-way Π2-alternating Turing machines cannotaccept unary nonregular languages in o(log n) space. This holdsfor an accept mode of space complexity measure, defined asthe worst cost of any accepting computation. This lower boundshould be compared with the corresponding bound for one-wayΣ2-alternating machines, that are able to accept unarynonregular languages in space O(log log n). Thus, Σ2-alternation is more powerful than Π2-alternationfor space bounded one-way machines with unary inputs.

We define L rational and L recognizable power series, and establish a Kleene-Schützenberger theorem for Lindenmayerian power series by showing that a power series is L rational if and only if it is L recognizable.

The aim of this paper is to show that a semi-commutation function can be expressed as the compound of a sequential transformation, a partial commutation function, and the reverse transformation. Moreover, we give a necessary and sufficient condition for the image of a regular language to be computed by the compound of two sequential functions and a partial commutation function.

We define a notion of asynchronous sliding block map that can be realized by transducers labeled in A* × B*. We show that, under some conditions, it is possible to synchronize this transducer by state splitting, in order to get a transducer which defines the same sliding block map and which is labeled in A × Bk, where k is a constant integer. In the case of a transducer with a strongly connected graph, the synchronization process can be considered as an implementation of an algorithm of Frougny and Sakarovitch for synchronization of rational relations of bounded delay. The algorithm can be applied in the case where the transducer has a constant integer transmission rate on cycles and has a strongly connected graph. It keeps the locality of the input automaton of the transducer. We show that the size of the sliding window of the synchronous local map grows linearly during the process, but that the size of the transducer is intrinsically exponential. In the case of non strongly connected graphs, the algorithm of Frougny and Sakarovitch does not keep the locality of the input automaton of the transducer. We give another algorithm to solve this case without losing the good dynamic properties that guaranty the state splitting process.