Semi-Supervised Robust Deep Neural Networks for Multi-Label Classification

In this paper, we propose a robust method for semisupervised training of deep neural networks for multi-label image classification. To this end, we use ramp loss, which is more robust against noisy and incomplete image labels compared to the classical hinge loss. The proposed method allows for learning from both labeled and unlabeled data in a semi-supervised learning setting. This is achieved by propagating labels from the labeled images to their unlabeled neighbors. Using a robust loss function becomes crucial here, as the initial label propagations may include many errors, which degrades the performance of non-robust loss functions. In contrast, the proposed robust ramp loss restricts extreme penalties for the samples with incorrect labels, and the label assignment improves in each iteration and contributes to the learning process. The proposed method achieves state-of-the-art results in semisupervised learning experiments on the CIFAR-10 and STL10 datasets, and comparable results to the state-of the-art in supervised learning experiments on the NUS-WIDE and MS-COCO datasets.