Quantization Impact on the Accuracy and Communication Efficiency Trade-off in Federated Learning for Aerospace Predictive Maintenance

Abdelkarim Loukili|April 9, 2026arXiv

Key Takeaway

For aerospace predictive maintenance on bandwidth-limited devices, 4-bit quantization (INT4) is the practical sweet spot—it cuts communication costs by 8× without sacrificing accuracy, while 2-bit quantization fails under realistic non-uniform data conditions despite appearing to work in simpl...

Summary

This paper studies how quantizing neural network gradients affects both accuracy and communication efficiency in federated learning for aircraft maintenance prediction.

efficiency training evaluation

Key Terms

quantization federated-learning non-iid gradient-communication communication-efficiency