Non-IID federated learning degrades because client loss landscapes diverge, and existing remedies work either by adjusting update paths or by modifying those landscapes.
In: ICLR (2021)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Understanding Federated Learning from IID to Non-IID dataset: An Experimental Study
Non-IID federated learning degrades because client loss landscapes diverge, and existing remedies work either by adjusting update paths or by modifying those landscapes.