Combined effects of client heterogeneity conditions

Determine the combined effect of quantity skew, missingness skew, and feature-distribution skew on Fed-ReMasker’s imputation performance under federated tabular feature-level missingness.

Background

The evaluation studies quantity skew, missingness skew, and feature-distribution skew separately in the heterogeneous-client experiments. Consequently, it does not establish how these forms of heterogeneity interact when they occur simultaneously, leaving their combined effect on Fed-ReMasker’s imputation performance unresolved.

References

The direction of the resulting performance difference is not known a priori: such heterogeneity could make collaborative learning either more difficult or, in some settings, more valuable.

— FedIncome: Federated Learning for Income Estimation in Digital Lending Under Data Sovereignty Constraints  (2609.27654 - Amed et al., 23 Sep 2026) in Section 6, Limitations and Further Research

We evaluate quantity, missingness, and feature-distribution skew separately, so their combined effect remains unknown.

— Fed-ReMasker: Federated Tabular Imputation under Feature-Level Missingness  (2609.28105 - Papathanail et al., 23 Sep 2026) in Section “Limitations and Future Work”