Isolate the independent contributions of TripleBound’s training components

Establish the independent effects of the CHGNN backbone, parser-inferred triplet supervision, communication-aware regularization, loss weighting, and removal of the original structure loss on microservice decomposition quality.

Background

TripleBound changes several aspects of the CHGNN baseline simultaneously, including triplet-based supervision, communication-aware regularization, loss weighting, and removal of the original structure loss. Because the reported comparison evaluates the complete configuration rather than incremental variants, the results cannot determine which modification produces the observed changes in structural modularity, inter-partition communication, or other metrics. An ablation study is therefore needed to disentangle the effects of the individual components.

References

The evaluation therefore establishes the behavior of the complete TripleBound configuration but cannot identify the independent contribution of each modification.

— TripleBound: Triplet-Guided Heterogeneous Graph Learning for Microservice Decomposition  (2609.11212 - Weerasinghe et al., 10 Sep 2026) in Section 6, “Threats to Validity,” subsection “Lack of component ablations”