Adaptive Gabor convolutional networks

作者：

Highlights：

• AGCNs use Gabor filters to manipulate convolutional kernels.

• We have proved that AGCNs can learn the invariant information from images.

• The parameters in the Gabor filters can be updated during the model training.

• We demonstrate the robustness of AGCNs against adversarial attacks.

摘要

•AGCNs use Gabor filters to manipulate convolutional kernels.•We have proved that AGCNs can learn the invariant information from images.•The parameters in the Gabor filters can be updated during the model training.•We demonstrate the robustness of AGCNs against adversarial attacks.

论文关键词：Gabor filters,Deep convolutional neural networks,Invariant information,Gabor convolutional filters,Image classification

论文评审过程：Received 31 March 2021, Revised 10 November 2021, Accepted 7 December 2021, Available online 15 December 2021, Version of Record 23 December 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108495