Learning Not to Learn in the Presence of Noisy Labels
This addresses the challenge of training robust models in the presence of mislabeled datasets, which is crucial for real-world applications where label noise is common.
The paper tackles the problem of learning with noisy labels by introducing the gambler's loss, which encourages models to abstain from learning on noisy data points, resulting in improved robustness and generalization across image and text classification tasks with strong results compared to baselines.
Learning in the presence of label noise is a challenging yet important task: it is crucial to design models that are robust in the presence of mislabeled datasets. In this paper, we discover that a new class of loss functions called the gambler's loss provides strong robustness to label noise across various levels of corruption. We show that training with this loss function encourages the model to "abstain" from learning on the data points with noisy labels, resulting in a simple and effective method to improve robustness and generalization. In addition, we propose two practical extensions of the method: 1) an analytical early stopping criterion to approximately stop training before the memorization of noisy labels, as well as 2) a heuristic for setting hyperparameters which do not require knowledge of the noise corruption rate. We demonstrate the effectiveness of our method by achieving strong results across three image and text classification tasks as compared to existing baselines.