A novel companion objective function for regularization of deep convolutional neural networks

作者：

Highlights：

• A novel companion objective function for training deep CNNs is proposed.

• We proposed two kinds of auxiliary supervision for deep CNNs.

• Auxiliary classifiers for hidden layers play a role as regularization.

• The companion objective function is compatible with other regularization strategies.

摘要

•A novel companion objective function for training deep CNNs is proposed.•We proposed two kinds of auxiliary supervision for deep CNNs.•Auxiliary classifiers for hidden layers play a role as regularization.•The companion objective function is compatible with other regularization strategies.

论文关键词：Convolutional neural network,Regularization,Companion objective function,Classification

论文评审过程：Received 15 May 2016, Revised 22 August 2016, Accepted 16 November 2016, Available online 24 November 2016, Version of Record 22 March 2017.

论文官网地址：https://doi.org/10.1016/j.imavis.2016.11.012