Improving neural network robustness through neighborhood preserving layers

作者：

Highlights：

• Propose a novel neighborhood preserving layer into neural network models.

• The proposed layer can replace fully-connected layers and are more robust against adversarial attack.

• Provide theoretical and experimental results to demonstrate the ad-vantage of our model

摘要

•Propose a novel neighborhood preserving layer into neural network models.•The proposed layer can replace fully-connected layers and are more robust against adversarial attack.•Provide theoretical and experimental results to demonstrate the ad-vantage of our model

论文关键词：Deep learning,Manifold approximation,Neighborhood preservation,Robustness,Adversarial attack,Image classification

论文评审过程：Received 30 January 2021, Revised 13 April 2022, Accepted 17 April 2022, Available online 1 May 2022, Version of Record 10 May 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104469