Adaptive Batch normalization

Hello everyone. Recently I went over the neural network pruning paper when I came across the term Adaptive Batch Normalization. The went over the Adaptive batch norm paper but I am not been able to understand completely. What I could gather was adaptive batch norm re-estimates the mean and variance on the target domain along with the train set. Am I understanding this wrong. Could anyone with a better understanding to the topic give a better explanation. Any help will be extremely appreciated.

© Copyright 2013-2020 Analytics Vidhya