CS Paper Radar

General semiring conditions for weighted L* termination

Noah Bertram, Daniel Luckhardt, Alexandra Silva

cs.FL cs.LO ·submitted 2026-10-30 03:03 UTC · v1

Abstract

Weighted automata learning has been well-studied over (semi)rings which have a dimension theory. In this work, we investigate active learning for weighted automata over semirings that lack such a structure. We first show that the principal ideal domain variant of the L* algorithm also terminates for a larger class of semirings containing Noetherian semirings, commutative rings, and bounded distributed lattices. We then present a variant of this algorithm for learning deterministic weighted automata, and prove that this algorithm terminates for commutative semirings.

Comment
22 pages

Closest papers

Nearest neighbours by meaning across the whole index.

Weighted Alternating Tree Automata

cs.FL ·2026-08-25 ·Olle Torstensson
published similarity 0.66
Abstract · In Proceedings AFL 2026, arXiv:2608.23071

We define weighted alternating tree automata — a generalization of weighted (non-alternating) tree automata, (unweighted) alternating tree automata, and weighted alternating (string) automata — over commutative semirings. We study their expressive power, and show that it is equivalent to that of weighted tree automata exactly when the employed semiring is locally finite. In our main result, we characterize the class of weighted tree languages defined by weighted alternating tree automata as the closure of the recognizable weighted tree languages under inverse tree homomorphisms.

0.46

The Value Generating Power of Weighted Tree Automata with Initial Algebra Semantics

cs.FL ·2026-08-25 ·Manfred Droste, Zoltán Fülöp, Andreja Tepavčević, Heiko Vogler
published similarity 0.60
Abstract · In Proceedings AFL 2026, arXiv:2608.23071

We consider the generating power of the initial algebra semantics of weighted tree automata over strong bimonoids (hence also over semirings) and the question under which conditions the weighted tree automata can produce only finitely many values. We show that there exists a right-distributive strong bimonoid which is bi-locally finite but not locally finite. We also show that if the ranked alphabet contains a symbol with rank at least two, then for any finitely generated strong bimonoid, weighted tree automata can generate, via their initial algebra semantics, all elements of the strong bimonoid. As a consequence of these results, for bi-locally finite right-distributive strong bimonoids which are not locally finite, weighted tree automata can generate infinitely many values, provided that the input ranked alphabet contains a symbol with rank at least two. This is in sharp contrast to the setting of weighted string automata, which can generate only finitely many values. As a further consequence, for any finitely generated semiring, there exists a weighted tree automaton which generates, via its run semantics, all elements of the semiring.

0.72

Learning Canonical Register Automata over Ordered Data Domains

cs.CL cs.LG ·2026-08-19 ·Yong Li, Qiyi Tang, Di-De Yen
similarity 0.58
Abstract

Register automata are finite automata equipped with memory that recognize data languages over infinite alphabets. In this work, we investigate active learning algorithms for deterministic register automata (DRAs) over ordered data domains--covering both dense domains, such as the rationals, and non-dense domains such as the integers. We show that the active learning problem for DRAs over both dense and non-dense ordered domains can be treated within a single unified framework. More specifically, we develop and implement a polynomial-time active learning procedure for DRAs over ordered domains, using oracles for membership, equivalence and memorability queries. The memorability queries were originally introduced for learning DRAs over domains with identity tests. Our unified framework also leads to a new consequence: minimization of DRAs over the non-dense ordered domain of integers is decidable, extending a result previously known only for dense domains. Finally, we give improved complexity bounds of several decision problems for DRAs over ordered domains that are closely related to the queries used in active learning.

0.27

Shortest Synchronizing Words on Product Automata

cs.FL ·2026-11-06 ·Henning Fernau, Igor Rystsov
similarity 0.56
Abstract

We consider some operations on finite automata graphs, in particular various product, union and merge operations, and investigate how they relate to the parameter~$\mathrm{rt}$, the \emph{length of a shortest synchronizing word}. We derive some lower and upper bounds for the operations that we study. We also examine several classes of automata (for instance, partially ordered, commutative or idempotent automata) that are preserved under these operations and show in which way the parameter $\mathrm{rt}$ behaves in these automata classes. Also, we consider automata classes that can be constructed by combining automata from better known classes; in particular, these new classes of automata will also satisfy the bound of the famous Černý conjecture. This also motivates us to study even more closure properties. In particular, we show that certain automata classes related to synchronization form pseudo-varieties. We also commence a similar study concerning the length of shortest homing sequences of Mealy automata.

0.13

Reliability-Aware Z-Grammars and Z-Automata over a Credible-Grade Semiring

cs.FL cs.LO ·2026-11-05 ·Sherzod Turaev, Nor Haniza Sarmin, Wan Heng Fong
theory similarity 0.56
Abstract · 35 pages, 7 figures, 3 tables

Fuzzy grammars and weighted automata usually attach to each word a single grade, which records how strongly the word is generated but not how far that record can be trusted. We study weighted grammars and finite automata whose weights are scalar grade-reliability pairs inspired by Z-numbers. Sequential composition multiplies both components, while aggregation selects a derivation by its credible grade, the product of grade and reliability, with ties resolved by grade. On the canonical carrier, in which the weights of zero grade or zero reliability are identified with zero, the resulting structure is a commutative selective semiring, isomorphic to a subsemiring of a lexicographically ordered product, and its credible-grade projection is a homomorphism onto the Viterbi semiring. We compare several alternative combination rules and identify the carriers on which the relevant algebraic laws hold. Standard results on weighted automata then yield a rational characterization and closure properties. We prove that positive credible-grade cuts and selected-run reliability filters of finite automata are regular and effectively constructible for rational weights and thresholds, and we exhibit distinctions that the scalar projection cannot see: a reliability filter that no scalar reweighting of the same one-state automaton implements, nonmonotonicity under the addition of runs, and an unambiguous automaton with no deterministic equivalent although its scalar projection is deterministic. A count-based example illustrates an attestation penalty for grammar rules. The construction concerns scalar pairs, and its extension to arbitrary fuzzy-valued Z-numbers remains open.

0.30

Graphoid Automata, Frobenius PROPs, and Semiring Evaluation

cs.FL math.CT ·2026-10-02 ·Antonios Kalampakas
theory similarity 0.54
Abstract · 29 pages

Graphoid automata evaluate graphs by composing relations between words of states. We examine the equations that make this evaluation independent of the chosen graph expression, and what changes when relations are replaced by matrices over a commutative semiring. We prove that the classical graphoid equations are equivalent to the special commutative Frobenius equations together with the required symmetry laws. The known cospan and hypergraph presentations then describe equality of structural and labelled expressions, with the bone law deleting precisely the isolated internal vertices. We use established partition representations to compare this equality with equality of evaluated matrices. For any nonempty finite state set $Q$ and commutative semiring $K$, we give a uniform criterion for two structural expressions to have the same matrix. The criterion determines which isolated-vertex counts are distinguished and when a vanishing scalar makes matrices from different boundary partitions equal. For the canonical structure, the bone law holds precisely when $|Q|\cdot1_K=1_K$. Preserving all relational compositions is a separate requirement, equivalent to additive idempotency for $|Q|\geq2$. We then examine the characteristic matrices of the known relational structures on $Q$ given by disjoint unions of Abelian groups. Their Frobenius equations hold over every commutative semiring, whereas specialness depends on the order of each group and the bone law depends on the number of groups. These conditions specify which equations remain valid when weights are assigned to a graphoid automaton.

0.27

Algebraic Decomposition Theory for Transformer Length Generalization

cs.FL cs.AI ·2026-08-13 ·Andy Yang, Blerta Veseli, Corentin Barloy, Michaël Cadilhac, Andreas Krebs +3 more
similarity 0.52
Abstract · 54 pages, 12 figures

Transformer-based language models are known to sometimes generalize to sequences longer than seen during training, but we lack a precise characterization of which tasks admit length generalization. It is not even known which regular languages transformers length-generalize on -- and this is a foundational class of languages. Our contributions are to establish the first complete characterization of which regular languages transformers length-generalize on and provide a decision algorithm running in polynomial time in the size of the language's syntactic monoid. These results rely on an effective characterization of the regular languages in C-RASP, a recently-established formalism that expresses which languages transformers length-generalize on. This characterization is challenging because classical tools like Krohn-Rhodes decomposition theory for finite semigroups are insufficient for C-RASP. Firstly, the basic building blocks of Krohn-Rhodes theory -- flip-flop and simple groups -- are not expressible in C-RASP. Secondly, the basic building block of C-RASP (unbounded counting) is not expressible by the finite semigroups of Krohn-Rhodes theory. Thus, length generalization on regular languages is controlled by an algebraic property that is invisible to classical finite decomposition theory. We generalize classical decomposition theory from finite semigroups to the infinite additive group on the integers, allowing us to characterize C-RASP in terms of iterated wreath products of the integers and derive a provable polynomial-time decision algorithm for regular language membership. Experiments across a broad test suite of regular languages confirm that our theory captures transformers' length-generalization behavior more accurately than existing classifications.

0.43

Proceedings 17th International Conference on Automata and Formal Languages

cs.FL cs.CC ·2026-08-24 ·Michal Hospodár, Galina Jirásková, Viktor Olejár, Juraj Šebej, György Vaszil
publishedhot direction similarity 0.51
Abstract

The 17th International Conference on Automata and Formal Languages (AFL 2026) was held in Košice, September 7-10, 2026. It was organized by the Faculty of Science, P. J. Šafárik University in Košice and the Mathematical Institute of the Slovak Academy of Sciences. This volume contains the texts of the three invited lectures and the 18 papers selected by the Program Committee from a total of 27 submissions. The accepted papers, submitted by authors from 14 countries, cover all important areas of automata and formal language theory and their applications such as grammars and acceptors for strings, trees, and graphs, algebraic theories and algorithms for automata and languages, decision problems, relations to complexity theory, picture description and analysis, and other fields.

0.53