Skip to main content
. 2021 Jul 19;11:14634. doi: 10.1038/s41598-021-93977-0

Table 3.

Test error (%) on the MNIST classification task of state-of-the-art methodologies based on ensemble over the years.

Method Year Test error (%)
Multi-column deep neural networks for image classification43 2012 0.23
Regularization of neural networks using dropconnect44 2013 0.21
RMDL:random multimodel deep learning for classification45 2018 0.18
Base-branching & Merging CNNw/HFCs20 2020 0.16
Efficient-CapsNet 2021 0.16