Contrast-Teacher-Student(CTS): A novel Semi-supervised Learning Approach for Sleep Staging

Abstract

Semi-supervised learning (SSL) can work well with limited labeled data and enormous unlabeled data. It is suitable for areas such as medical treatment or biological research, whose data is sufficient but the label is high-cost. Unfortunately, few researchers are doing relevant studies. In this paper, we propose Contrast-Teacher- Student(CTS) model, a novel semi-supervised learning approach for physiological signal processing which is important in the diagnosis of some diseases. Our method achieves good performance with only a small amount of labeled data, and our method outperforms the current mainstream semi-supervised learning methods on the public dataset.