Skip to main content

Table 2 Tukey’s HSD test with groupings

From: A parallel and distributed stochastic gradient descent implementation using commodity clusters

 

Training time

Groups

16

71.06

a

8

36.74

b

1

22.82

c

4

22.78

c

2

21.98

c

  1. The slowest is the 16 node cluster, second slowest is the 8 node cluster, and all clusters in group “C” have no difference in training time
  2. Training time vs. # nodes, 1:1