Table 5 Scores of different number of clients in federated learning (averaged over 5 repetitions).

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

Scenario

Results

Clients

Train patches

Validation patches

Precision

Recall

F1-W

F1-W-NB

F1 Macro

2

2156

240

0.917

0.923

0.916

0.819

0.659

4

1078

120

0.913

0.919

0.913

0.808

0.63

8

539

60

0.905

0.91

0.904

0.785

0.576

16

270

30

0.877

0.897

0.886

0.739

0.456

32

135

15

0.863

0.886

0.874

0.713

0.435

64

67

7

0.844

0.866

0.854

0.662

0.399

128

34

4

0.757

0.817

0.783

0.442

0.258

256

17

2

0.616

0.747

0.667

0.1

0.108

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