Knowledge distilling based model compression and feature learning in fault diagnosis

Wenfeng Zhang, Gautam Biswas, Qi Zhao, Hongbo Zhao, Wenquan Feng

Abstract: Highlights•An occlusion-based importance analysis method for significant input variables and learned features selection.•The student network design relies on the simplified network topology based on the importance analysis.•A knowledge distilling approach for model compression and important feature learning.