An $L^{\#}$ Based Algorithm for Active Learning of Minimal Separating Automata

TL;DR

The paper introduces a new active learning algorithm based on $L^{ ext{ iny extnumero}}$ for efficiently learning minimal separating DFAs between two disjoint languages, with superior performance demonstrated on various benchmarks.

Contribution

It presents a novel $L^{ ext{ iny extnumero}}$-based algorithm for active learning of minimal separating automata, improving over existing methods.

Findings

Significantly outperforms existing algorithms on benchmarks

Successfully learns minimal separating DFAs when they exist

Effective on both synthetic and industrial data

Abstract

A DFA separates two disjoint languages $L_1$ and $L_2$ if it accepts every word in $L_1$ and rejects every word in $L_2$. Algorithms for active learning of small separating DFAs have many applications, e.g., for learning network invariants, learning contextual assumptions in compositional verification, learning state machines from large amounts of log data, and learning bug pattern descriptions. We propose a simple active learning algorithm, inspired by $L^{\#}$, that learns a minimal separating DFA for disjoint languages $L_1$ and $L_2$ if one exists. Experiments show that our algorithm significantly outperforms existing active learning algorithms on both randomly generated and industrial benchmarks.