Two-wayness: Automata and Transducers

This PhD is about two natural extensions of Finite Automata (FA): the 2-way fa (2FA) and the 2-way transducers (2T). It is well known that 2FA s are computably equivalent to FAs, even in their nondeterministic (2nfa) variant. However, in the field of descriptional complexity, some questions remain. Raised by Sakoda and Sipser in 1978, the question of the cost of the simulation of 2NFA by 2DFA (the deterministic variant of 2FA) is still open. In this manuscript, we give an answer in a restricted case in which the nondeterministic choices of the simulated 2NFA may occur at the boundaries of the input tape only (2ONFA). We show that every 2ONFA can be simulated by a 2DFA of subexponential (but superpolynomial) size. Under the assumptions L=NL, this cost is reduced to the polynomial level. Moreover, we prove that the complementation and the simulation by a halting 2ONFA is polynomial. We also consider the anologous simulations for alternating devices. Providing a one-way write-only output tape to FAs leads to the notion of transducer. Contrary to the case of finite automata which are acceptor, 2-way transducers strictly extends the computational power of 1-way one, even in the case where both the input and output alphabets are unary. Though 1-way transducers enjoy nice properties and characterizations (algebraic, logical, etc. . . ), 2-way variants are less known, especially the nondeterministic case. In this area, this manuscript gives a new contribution: an algebraic characterization of the relations accepted by two-way transducers when both the input and output alphabets are unary. Actually, it can be reformulated as follows: each unary two-way transducer is equivalent to a sweeping (and even rotating) transducer. We also show that the assumptions made on the size of the alphabets are required, that is, sweeping transducers weakens the 2-way transducers whenever at least one of the alphabet is non-unary. On the path, we discuss on the computational power of some algebraic operations on word relations, introduced in the aim of describing the behavior of 2-way transducers or, more generally, of 2-way weighted automata. In particular, the mirror operation, consisting in reversing the input word in order to describe a right to left scan, draws our attention. Finally, we study another kind of operations, more adapted for binary word relations: the composition. We consider the transitive closure of relations. When the relation belongs to some very restricted sub-family of rational relations, we are able to compute its transitive closure and we set its complexity. This quickly becomes uncomputable when higher classes are considered.