Skip to main content
. 2020 Jan 21;10:834. doi: 10.1038/s41598-020-57866-2

Figure 3.

Figure 3

Histogram showing the distribution of the classification error from all iterated network structures in each subset of the MNIST dataset (100, 500 and 1000 samples). The three distributions are similar regardless of sample size, suggesting that the sample size contributes less to the optimal network structure search procedure.