Skip to main content

Table 2 The convergence rates on approximation methods LA and GH. (Both LA and GH held the same convergence threshold \(10^{-3}\). The mean values and standard deviations (in parentheses) were given)

From: Federated learning algorithms for generalized mixed-effects model (GLMM) on horizontally partitioned data from distributed sources

Setting

LA

GH

Steps

Runtime (s)

Steps

Runtime (s)

1

22.875 (21.623)

47.953 (20.513)

34.850 (9.213)

104.460 (10.614)

2

21.500 (21.977)

40.947 (36.466)

35.000 (8.711)

100.940 (19.940)

3

29.867 (31.719)

108.931 (65.486)

34.900 (6.138)

1259.285 (231.956)

4

27.846 (24.034)

84.343 (76.502)

36.650 (6.310)

1342.695 (250.603)

5

59.722 (42.057)

10.631 (3.945)

33.750 (10.146)

12.568 (2.116)

6

67.188 (48.994)

10.499 (4.054)

31.400 (11.081)

11.430 (3.064)

7

96.286 (53.635)

96.501 (38.632)

37.450 (3.818)

369.165 (41.998)

8

116.083 (46.479)

91.304 (62.410)

37.150 (4.295)

309.693 (36.621)