MILD: Modeling the Instance Learning Dynamics for Learning with Noisy Labels
This work addresses the challenge of reducing network memorization on falsely-labeled data for noisy-label learning, which is important for applications where accurate labels are costly, but it appears incremental as it builds on prior small-loss heuristics.
The paper tackles the problem of learning with noisy labels by proposing an iterative selection method based on the Weibull mixture model to identify clean data using instance learning dynamics, and it outperforms existing methods in experiments on synthetic and real-world datasets.
Despite deep learning has achieved great success, it often relies on a large amount of training data with accurate labels, which are expensive and time-consuming to collect. A prominent direction to reduce the cost is to learn with noisy labels, which are ubiquitous in the real-world applications. A critical challenge for such a learning task is to reduce the effect of network memorization on the falsely-labeled data. In this work, we propose an iterative selection approach based on the Weibull mixture model, which identifies clean data by considering the overall learning dynamics of each data instance. In contrast to the previous small-loss heuristics, we leverage the observation that deep network is easy to memorize and hard to forget clean data. In particular, we measure the difficulty of memorization and forgetting for each instance via the transition times between being misclassified and being memorized in training, and integrate them into a novel metric for selection. Based on the proposed metric, we retain a subset of identified clean data and repeat the selection procedure to iteratively refine the clean subset, which is finally used for model training. To validate our method, we perform extensive experiments on synthetic noisy datasets and real-world web data, and our strategy outperforms existing noisy-label learning methods.