Continuity of Functional Transducers: A Profinite Study of Rational Functions

On the Circuit Complexity of Sigmoid Feedforward Neural Networks

We investigate the intersection problem for finite semigroups, which asks for a given set of regular languages, represented by recognizing morphisms to finite semigroups, whether there exists a word contained in their intersection. We introduce compressibility measures as a useful tool to classify the intersection problem for certain classes of finite semigroups into circuit complexity classes and Turing machine complexity classes. Using this framework, we obtain a new and simple proof that for groups and commutative semigroups, the problem is contained in \(\mathsf {NP}\). We uncover certain structural and non-structural properties determining the complexity of the intersection problem for varieties of semigroups containing only trivial submonoids. More specifically, we prove \(\mathsf {NP}\)-hardness for classes of semigroups having a property called unbounded order and for the class of all nilpotent semigroups of bounded order. On the contrary, we show that bounded order and commutativity imply containment in the circuit complexity class \(\mathsf {qAC}^k\) (for some \(k \in \mathbb {N}\)) and decidability in quasi-polynomial time. We also establish connections to the monoid variant of the problem.

We show that the class of string languages generated by linear context-free rewriting systems is equal to the class of output languages of deterministic tree-walking transducers. From equivalences that have previously been established we know that this class of languages is also equal to the string languages generated by context-free hypergraph grammars, multicomponent tree-adjoining grammars, and multiple context-free grammars and to the class of yields of images of the regular tree languages under finite-copying top-down tree transducers.

It is known that both the class of all (n,k)-languages and the class of all time-variant languages over a finite alphabet contain the class of all regular languages. In this paper we show that in general neither one of these two classes of languages contains the other by constructing an infinite sequence of strongly primitive words over an alphabet with four letters.

This paper examines the circuit complexity of feedforward neural networks having sigmoid activation function. The starting point is the complexity class NN k defined in [18]. First two additional complexity classes N N Δ k https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315784076/1ea703af-0c66-41e4-8ff3-9e9c700ac445/content/eq1772.tif/> and N N Δ , e k https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315784076/1ea703af-0c66-41e4-8ff3-9e9c700ac445/content/eq1773.tif/> having less restrictive conditions (than NN k ) concerning fan-in and accuracy are defined. We then prove several relations among these three classes and well established circuit complexity classes. All the proofs are constructive as being built around binary adders. By relaxing the fan-in condition from logarithmic [18] to (almost) polynomial, the results show that substantially larger classes of sigmoid activation feedforward neural networks can be implemented in polynomial size Boolean circuits. This is done on the expense of a logarithmic factor increase in the number of layers, but having constant fan-in (of 2). The main conclusion is that there are interesting fan-in dependent depth-size tradeoffs when trying to digitally implement sigmoid activation neural networks.

In this paper, we define three classes of languages of bi-infinite words, namely local bi-ω languages, recognizable bi-ω languages and Buchi local bi-ω languages as subclasses of the class of regular bi-ω languages and prove some basic results. We observe that the class of recognizable bi-ω languages coincides with the well-known class of rational bi-adherence languages and show that the class of regular bi-ω languages is the class of morphic images of Buchi local bi-ω languages. We provide learning algorithms for Buchi local bi-ω languages and regular bi-ω languages.

We present the first polynomial time algorithm to learn nontrivial classes of languages of infinite trees. Specifically, our algorithm uses membership and equivalence queries to learn classes of omega-tree languages derived from weak regular omega-word languages in polynomial time. The method is a general polynomial time reduction of learning a class of derived omega-tree languages to learning the underlying class of omega-word languages, for any class of omega-word languages recognized by a deterministic Büchi acceptor. Our reduction, combined with the polynomial time learning algorithm of Maler and Pnueli [Maler and Pneuli, Inform. Comput., 1995] for the class of weak regular omega-word languages yields the main result. We also show that subset queries that return counterexamples can be implemented in polynomial time using subset queries that return no counterexamples for deterministic or non-deterministic finite word acceptors, and deterministic or non-deterministic Büchi omega-word acceptors. A previous claim of an algorithm to learn regular omega-trees due to Jayasrirani, Begam and Thomas [Jayasrirani et al., ICGI, 2008] is unfortunately incorrect, as shown in [Angluin, YALEU/DCS/TR-1528, 2016].

Two kinds of automata are presented, for recognising new classes of regular and context-free nominal languages. We compare their expressive power with analogous proposals in the literature, showing that they express novel classes of languages. Although many properties of classical languages hold no longer in the nominal case, we design a slight restriction of our models that preserve some interesting ones. In particular, we prove the emptiness problem decidable and we construct the intersection between (restricted) regular and context-free automata. By examples and walking through their properties we argue the relevance of our models in the context of the verification of resource usage patterns.

This paper studies the relaxation of word-order, using dependency grammars (DG’s). We study ‘relax-ability of languages (and DG’s)’ based on the notion of DR-(non)projectivity. We obtain two infinite scales of classes of languages using different degrees of relaxation. The one scale contains the class of context-free languages as its first (minimal) element, the other one starts by the class of regular languages.

Closure properties of linear context-free tree languages with an application to optimality theory

We characterize the classes of languages described by P automata, i.e., accepting P systems with communication rules only. Motivated by properties of natural computing systems, we study computational complexity classes with a certain restriction on the use of the available workspace in the course of computations and relate these to the language classes described by P automata. We prove that if the rules of the P system are applied sequentially, then the accepted language class is strictly included in the class of languages accepted by one-way Turing machines with a logarithmically bounded workspace, and if the rules are applied in the maximal parallel manner, then the class of context-sensitive languages is obtained.KeywordsTuring MachineMathematical LinguisticsRule ApplicationInput SymbolInput TapeThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Learning approximately regular languages with reversible languages

Separations among the first order logic Ring(0,+, ∗) of finite residue class rings, its extensions with generalized quantifiers, and in the presence of a built-in order are shown, using algebraic methods from class field theory. These methods include classification of spectra of sentences over finite residue classes as systems of congruences, and the study of their h-densities over the set of all prime numbers, for various functions h on the natural numbers. Over ordered structures the logic of finite residue class rings and extensions are known to capture DLOGTIMEuniform circuit complexity classes ranging from AC to TC. Separating these circuit complexity classes is directly related to classifying the h-density of spectra of sentences in the corresponding logics of finite residue classes. We further give general conditions under which a logic over the finite residue class rings has a sentence whose spectrum has no h-density. One application of this result is that Supported by MICINN project TIN2011-27479-C04-03 (BASMATI), MINECO project TIN201457226-P (APCOM) and Gen. Cat. project SGR2014-890 (MACDA).

Model checking is often cited as a success story for transitioning and engineering ideas rooted in logics and automata to practice. In this paper, we discuss how the efforts aimed at improving the scope and effectiveness of model checking tools have revived the study of logics and automata leading to unexpected theoretical advances whose impact is not limited to model checking. In particular, we describe how our efforts to add context-free specifications to software model checking led us to the model of nested words as a representation of data with both a linear ordering and a hierarchically nested matching of items. Such dual structure occurs in diverse corners of computer science ranging from executions of structured programs where there is a well-nested matching of entries to and exits from functions and procedures, to XML documents with the hierarchical structure specified by start-tags matched with end-tags. Finite-state acceptors of nested words define the class of regular languages of nested words that has all the appealing theoretical properties that the class of regular word languages enjoys. We review the emerging theory of nested words, its extension to nested trees, and its potential applications.

An operation of concatenation is defined for graphs. This allows strings to be viewed as expressions denoting graphs, and string languages to be interpreted as graph languages. For a class \(K\) of string languages, \({\rm Int}(K)\) is the class of all graph languages that are interpretations of languages from \(K\). For the classes REG and LIN of regular and linear context-free languages, respectively, \({\rm Int}({\rm REG}) = {\rm Int}({\rm LIN})\). \({\rm Int}({\rm REG})\) is the smallest class of graph languages containing all singletons and closed under union, concatenation and star (of graph languages). \({\rm Int}({\rm REG})\) equals the class of graph languages generated by linear HR (= Hyperedge Replacement) grammars, and \({\rm Int}(K)\) is generated by the corresponding \(K\)-controlled grammars. Two characterizations are given of the largest class \(K'\) such that \({\rm Int}(K') = {\rm Int}(K)\). For the class CF of context-free languages, \({\rm Int}({\rm CF})\) lies properly inbetween \({\rm Int}({\rm REG})\) and the class of graph languages generated by HR grammars. The concatenation operation on graphs combines nicely with the sum operation on graphs. The class of context-free (or equational) graph languages, with respect to these two operations, is the class of graph languages generated by HR grammars.