Identify SUL classes suited for output-based component decomposition (Duhaiby & Groce, ICSE 2020)

Determine the classes of systems under learning (SULs) for which the output-based component decomposition algorithm of Duhaiby and Groce (ICSE 2020) is applicable and effective. Precisely characterize structural or behavioral properties of Mealy/Moore systems or labeled transition systems that make the algorithm suitable, and delineate cases where it does not yield advantages over monolithic learning.

Background

The paper surveys several compositional automata learning approaches and contrasts them with the authors’ componentwise setting. In discussing Duhaiby & Groce (ICSE 2020), the authors note that their method decomposes systems according to output, relaxing earlier disjointness assumptions. However, the algorithm requires certain graph-theoretic conditions on the target system that may be nontrivial to verify.

Against this backdrop, the authors explicitly acknowledge that it remains undetermined which SULs are most appropriate for Duhaiby & Groce’s approach. Clarifying this would guide practitioners on when to apply output-based decomposition and help benchmark its performance relative to monolithic or input-decomposition strategies.