Revealing the influence of participant failures on model quality in cross-silo Federated Learning

Researchers conducted extensive experiments to analyze how participant failures affect Federated Learning model quality across different data types and scenarios. The study reveals that data skewness significantly impacts model performance and can lead to overly optimistic evaluations when participants are missing from the training process.

• →Federated Learning systems are susceptible to crash failures and network partitioning that can compromise model reliability.
• →Data skewness has a strong impact on model performance when participants are missing from FL training.
• →Missing participants can lead to overly optimistic model evaluations in certain scenarios.
• →The study examined various factors including data types, availability patterns, and model architectures across image, tabular, and time-series data.
• →Reliability issues in FL can compromise the validity, stability, and reproducibility of machine learning outcomes.