Crowdsourced federated learning with inconsistent label representation

Key Points

• Inconsistent label representation complicates federated learning outcomes, affecting algorithm efficacy.
• Data from diverse sources demonstrates variability in labeling, highlighting the need for standardization.
• The analysis critiques current algorithms in federated learning, emphasizing their limitations in data privacy and accuracy.
• Improving label consistency may enhance the performance of crowdsourced systems, warranting further exploration.