Correlation-based structural dropout for convolutional neural networks

作者：

Highlights：

• CorrDrop regularizes CNNs by dropping feature units based on feature correlation.

• A structural dropout method can effectively drop features in CNNs.

• Spatial-wise and channel-wise CorrDrop are proposed.

• Extensive experiments show the superiority of CorrDrop over other counterparts.

摘要

•CorrDrop regularizes CNNs by dropping feature units based on feature correlation.•A structural dropout method can effectively drop features in CNNs.•Spatial-wise and channel-wise CorrDrop are proposed.•Extensive experiments show the superiority of CorrDrop over other counterparts.

论文关键词：Over-fitting,Regularization,Dropout,Convolutional neural networks

论文评审过程：Received 9 May 2020, Revised 25 May 2021, Accepted 14 June 2021, Available online 16 June 2021, Version of Record 26 June 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108117