Three-teaching: A three-way decision framework to handle noisy labels

Guoqing Chao, Kaiwen Zhang, Xiru Wang, Dianhui Chu

Abstract: Highlights•We proposed a three-teaching model which introduces another neural network and “voting mechanism” to guarantee the prediction quality while facing noisy labels.•Three-teaching performs better than co-teaching and other compared methods on several real-world datasets.•Three-teaching provides a new way to implement three-way decision.