Learning competitive channel-wise attention in residual network with masked regularization and signal boosting
作者:
Highlights:
• Some strategies are proposed to jointly model the relationship of residual and identity channels.
• A masked regularization and a signal boosting method are proposed to improve diversity of features.
• These methods can be cheaply applied to residual networks.
摘要
•Some strategies are proposed to jointly model the relationship of residual and identity channels.•A masked regularization and a signal boosting method are proposed to improve diversity of features.•These methods can be cheaply applied to residual networks.
论文关键词:Residual networks,Competitive channel-wise attention,Masked regularization,Signal boosting
论文评审过程:Received 10 May 2019, Revised 8 February 2020, Accepted 21 May 2020, Available online 30 May 2020, Version of Record 29 June 2020.
论文官网地址:https://doi.org/10.1016/j.eswa.2020.113591