Robust Training with Ensemble Consensus
This addresses label noise in deep learning, which can degrade model generalization, but it is incremental as it builds on existing ensemble and noise-handling techniques.
The paper tackles the problem of deep neural networks memorizing noisy labels by proposing Learning with Ensemble Consensus (LEC), a training method that removes noisy examples based on ensemble consensus, achieving state-of-the-art performance on noisy MNIST, CIFAR-10, and CIFAR-100 datasets.
Since deep neural networks are over-parameterized, they can memorize noisy examples. We address such a memorization issue in the presence of label noise. From the fact that deep neural networks cannot generalize to neighborhoods of memorized features, we hypothesize that noisy examples do not consistently incur small losses on the network under a certain perturbation. Based on this, we propose a novel training method called Learning with Ensemble Consensus (LEC) that prevents overfitting to noisy examples by removing them based on the consensus of an ensemble of perturbed networks. One of the proposed LECs, LTEC outperforms the current state-of-the-art methods on noisy MNIST, CIFAR-10, and CIFAR-100 in an efficient manner.