We present here theoretical results coming from the implementation of the package called AMULT (automata with multiplicities in several noncommutative variables). We show that classical formulas are "almost every time" optimal, characterize the dual laws preserving rationality and also relators that are compatible with these laws. boolean coefficients). In 1974, for the case of fields, Fliess [6] extended the proof of the equivalence of minimal linear representations, using Hankel matrices. All these results allow us to construct an algorithmic processing for this series and their associated operations. In fact, classical constructions of language theory have multiplicity analogues which can be used in every domain where linear recurrences between words are handled. All these operations can be found in the package over automata with multiplicities (called AMULT). This package is a component of the environment SEA (Symbolic Environment for Automata) under development at the University of Rouen.The structure of this paper is the following: In section 3 (the first section after introductory paragraphs), we recall the classical construction for simple rational laws (+, ., * , ×) and make some remarks concerning in particular the non-commutative case. The compositions are based on polynomial formulas which has an important consequence on composition of automata choosen "at random". In fact, this first result says that the classical formulas are "almost everywhere" optimal (which is clear from experimental tests at random).In section 4, we show that the three laws known to preserve rationality ( Hadamard, shuffle and infiltration products) are of the same nature: they arise by dualizing alphabetic morphisms. Moreover, they are, up to a deformation, the only ones of this kind, which of course, shows immediately in the implemented formulas.Section 5 is devoted to study the compatibility with relators. It was well known that, when coefficients are taken in a ring of characteristic 0, the only relators compatible with the shuffle were partial commutations ([3]). Here, we show that a similar result holds (up to the supplementary possibility of letters erasure) when K is a semiring which is not a ring. This implies the known case as a corollary. To end with, we give examples of some strange relators in characteristic 2. PreambleLet K A be the set of noncommutative formal series with A a finite alphabet and K a semiring (commutative or not). A series denoted S = w∈A * S|w w is recognizable iff there exists a row vector λ ∈ K 1×n , a morphism of monoids µ : A * → K n×n and a column vector γ ∈ K n×1 , such that for all w ∈ A * , one has S|w = λµ(w)γ. Throughout the paper, we will denote by S : (λ, µ, γ) this property and say that (λ, µ, γ) is a linear repre-
We take an active interest in the problem of conversion of a Weighted Finite Automaton (WFA) into a K-expression. The known algorithms give an exponential size expression in the number of states of the given automaton. We study the McNaughton-Yamada algorithm in the case of multiplicities and then we show that the resulting K-expression is in the Star Normal Form (SNF) defined by Brüggemann-Klein [3].The Glushkov algorithm computes an (n + 1)-state automaton from an expression having n occurrences of letters even in the multiplicity case [5]. We reverse this procedure and get a linear size K-expression from a Glushkov WFA. A characterization of Glushkov WFAs which are not in SNF is given. This characterization allows us to emphasize a normal form for K-expressions. As for SNF in the boolean case, we show that every K-expression has an equivalent one in normal form having the same Glushkov WFA. We end with an algorithm giving a small normal form K-expression from a Glushkov WFA.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.