Synergistic Network Learning and Label Correction for Noise-robust Image Classification

Large training datasets almost always contain examples with inaccurate or incorrect labels. Deep Neural Networks (DNNs) tend to overfit training label noise, resulting in poorer model performance in practice. To address this problem, we propose a robust label correction framework combining the ideas of small loss selection and noise correction, which learns network parameters and reassigns ground true labels iteratively. Taking the expertise of deep neural networks to learn meaningful patterns before fitting noise, our framework first trains two networks over the current dataset with small loss selection in each round. Based on the classification loss and agreement loss of two networks, we can measure the confidence of training data. More and more confident samples then can be selected for label correction during the learning process. We demonstrate our method on both synthetic and real-world datasets with different noise types and noise rates, including MNIST, CIFAR-10, CIFAR-100 and Clothing1M, where our method outperforms many state-of-the-art approaches.