Table 5:

The total runtime in seconds to train central and federated models, averaged over K folds. Algorithms are grouped by aggregation strategy (Central, FedAvg, FedProx). The lowest training time for each group is bolded.

Algorithm Name	Runtime (Seconds)
LogisticRegression	1.857e-02 ± 0.008
SGDClassifier	6.771e-03 ± 0.001
MLPClassifier	1.609e-01 ± 0.009
XGBRFClassifier	1.633e-01 ± 0.003
FedAvg SGDClassifier	1.513e+01 ± 1.497
FedAvg XGBRFClassifier	1.061e+01 ± 0.014
FedAvg LRClassifier	7.909e+00 ± 0.550
FedAvg MLPClassifier	8.755e+00 ± 0.141
FedProx μ = 0.5 LRClassifier	8.747e+00 ± 0.158
FedProx μ = 0.5 MLPClassifier	9.039e+00 ± 0.266
FedProx μ = 2 LRClassifier	8.905e+00 ± 0.130
FedProx μ = 2 MLPClassifier	9.260e+00 ± 0.163