no code implementations • 18 Aug 2020 • Divya Gaur, Joachim Folz, Andreas Dengel
The main purpose of this work is to determine if it is possible to train networks effectively when batch normalization is removed through adaption of the training process.
