Abstract
Since deep neural networks are over-parametrized, they may memorize noisy examples. We address such memorizing issue in the presence of annotation noise. From the fact that deep neural networks cannot generalize neighborhoods of the features acquired via memorization, we observe that noisy examples do not consistently incur small losses on the network under a certain perturbation. Based on this observation, we propose a novel training method called Learning with Ensemble Consensus (LEC) that prevents overfitting noisy examples by eliminating them using the consensus of an ensemble of perturbed networks. One of the proposed LECs, LTEC outperforms the current state-of-the-art methods on MNIST, CIFAR-10, and CIFAR-100 with efficient memory usage.