Table 2 Hyperparameter study for centralised model.

From: Balancing centralisation and decentralisation in federated learning for Earth Observation-based agricultural predictions

 

No threshold

0.5 Threshold

0.75 Threshold

Loss ratio (\(\lambda\))

F1-W

F1-W-NB

F1-W

F1-W-NB

F1-W

F1-W-NB

0.1

0.875

0.725

0.875

0.725

0.870

0.713

0.3

0.883

0.741

0.882

0.739

0.870

0.710

0.5

0.900

0.766

0.898

0.761

0.887

0.732

0.7

0.898

0.779

0.895

0.773

0.860

0.690

0.9

0.897

0.778

0.895

0.775

0.873

0.724

  1. Significant values are in bold.
Back to article page