Self-paced resistance learning against overfitting on noisy labels

作者：

Highlights：

• We propose a novel resistance loss to alleviate model overfitting.

• We propose a novel framework to robustly train CNNs on noisy labels.

• It integrates CNNs’ memorization effect, curriculum learning with a resistance loss.

摘要

•We propose a novel resistance loss to alleviate model overfitting.•We propose a novel framework to robustly train CNNs on noisy labels.•It integrates CNNs’ memorization effect, curriculum learning with a resistance loss.

论文关键词：Convolutional neural networks,Self-paced resistance,Model overfitting,Noisy labels

论文评审过程：Received 14 May 2022, Revised 21 July 2022, Accepted 20 September 2022, Available online 29 September 2022, Version of Record 4 October 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109080