Data cleaning and classification in the presence of label noise with class-specific autoencoder

Abstract

We present a simple but effective method for data cleaning and classification in the presence of label noise. The key idea is to treat the data points with label noise as outliers of the class indicated by the corresponding noisy label. However, finding such dubious observations is challenging in general. We therefore propose to reduce their potential influence using feature learning method by class-specific autoencoder. Particularly, we learn for each class a feature space using all the samples labeled as that class, including those with noisy labels. Furthermore, in the case of high label noise, we propose a weighted class-specific autoencoder by considering the effect of each data point. To fully exploit the advantage of the learned feature space, we use a minimum reconstruction error based method for testing. Experiments on several datasets show that the proposed method achieves state-of-the-art performance on the related tasks with noisy labels.