arXiv ScienceSearch

arXiv · 2411.10601

Learning Quantitative Automata Modulo Theories

Abstract

Quantitative automata are useful representations for numerous applications, including modeling probability distributions over sequences to Markov chains and reward machines. Actively learning such automata typically occurs using explicitly gathered input-output examples under adaptations of the L-star algorithm. However, obtaining explicit input-output pairs can be expensive, and there exist scenarios, including preference-based learning or learning from rankings, where providing constraints is a less exerting and a more natural way to concisely describe desired properties. Consequently, we propose the problem of learning deterministic quantitative automata from sets of constraints over the valuations of input sequences. We present QUINTIC, an active learning algorithm, wherein the learner infers a valid automaton through deductive reasoning, by applying a theory to a set of currently available constraints and an assumed preference model and quantitative automaton class. QUINTIC performs a complete search over the space of automata, and is guaranteed to be minimal and correctly terminate. Our evaluations utilize theory of rationals in order to learn summation, discounted summation, product, and classification quantitative automata, and indicate QUINTIC is effective at learning these types of automata.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Eric Hsiung, Swarat Chaudhuri, Joydeep Biswas. 2024-11-15. Learning Quantitative Automata Modulo Theories. https://arxiv.org/abs/2411.10601

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Regular Expressions with Backreferences on Multiple Context-Free Languages, and the Closed-Star Condition

Backreference is a well-known practical extension of regular expressions and is supported by the regular expression engines in the standard libraries of most modern programming languages, such as Java, Python, JavaScript and more. A difficulty of backreference is non-regularity: backreference strictly enhances the expressive power of regular expressions to the point that regular expressions with backreferences (rewbs) can describe non-regular (in fact, even non-context-free) languages. In this paper, we investigate the expressive power of rewbs by comparing rewbs to multiple context-free languages (MCFL) and parallel multiple context-free languages (PMCFL). First, we prove that the language class of rewbs is a proper subclass of unary-PMCFLs, which coincide with the EDT0L languages. Our result strictly improves the known (non-trivial) upper bound of rewbs, because the best-known bound was the intersection of the class of nondeterministic logspace languages and that of indexed languages, and, as we shall show in this paper, the class of EDT0L languages is a proper subclass of the intersection. Additionally, we show that, however, the language class of rewbs is not contained in that of MCFLs even when restricted to rewbs with only one capturing group and no captured references. Therefore, in general, the parallelism seems essential for rewbs. Backed by these results, we define a novel syntactic condition on rewbs that we call closed-star and observe that it provides an upper bound on the number of times a rewb references the same captured string. The closed-star condition allows dispensing with the parallelism: we prove that the language class of closed-star rewbs falls inside the class of unary-MCFLs, which is equivalent to that of EDT0L systems of finite index. Furthermore, we show that the language class of closed-star rewbs also falls inside the class of nonerasing stack languages.

cs.FL

Compressed Subsequence Checking is PSPACE-complete

It is shown that the (scattered) subsequence problem for two words represented by straight-line programs is PSPACE-complete, even over a binary alphabet. The lower bound is obtained by a polynomial-time reduction from quantified subset sum.

cs.FL

On the Kanazawa--Salvati Conjecture

The language $\mathrm{MIX}$ consists of all words over a three-letter alphabet that have an equal number of occurrences of each letter. It is also the word problem of $\mathbb{Z}^2$ with respect to a suitable choice of generators. The Kanazawa--Salvati conjecture states that $\mathrm{MIX}$ is not a well-nested multiple context-free language. Every well-nested multiple context-free language is an indexed language. We reduce the conjecture to an explicit combinatorial problem about tuples of words, which is easier to state than the original formulation in terms of arbitrary well-nested multiple context-free grammars. More generally, for every surjective monoid homomorphism $ψ\colon Σ^* \to \mathbb{Z}^d$, we define a family of well-nested multiple context-free grammars $G_ψ[r]$ for $r \geq 1$, each of which generates a sublanguage of $ψ^{-1}(\mathbf{0})$. We prove that every well-nested multiple context-free sublanguage of $ψ^{-1}(\mathbf{0})$ is contained in $L(G_ψ[r])$ for some $r \geq 1$. Using this family, we prove that the four-letter analogue $\mathrm{MIX}_4$, which is a word problem of $\mathbb{Z}^3$, is not a well-nested multiple context-free language. The proof reduces this claim to a result of Bishop--Elder--Evetts--Gallot--Levine stating that the two-letter analogue $\mathrm{MIX}_2$ is not generated by any non-branching multiple context-free grammar. The Kanazawa--Salvati conjecture itself remains open.

cs.FL