CTRL: Clustering Training Losses for Label Error Detection
This addresses the issue of corrupted labels in supervised learning datasets, which can degrade model generalization, by providing an effective detection method, though it appears incremental as it builds on existing observations about learning differences.
The paper tackles the problem of detecting label errors in multi-class datasets by proposing CTRL, a framework that clusters training losses to distinguish clean and noisy labels, achieving state-of-the-art error detection accuracy on image and tabular datasets under simulated noise.
In supervised machine learning, use of correct labels is extremely important to ensure high accuracy. Unfortunately, most datasets contain corrupted labels. Machine learning models trained on such datasets do not generalize well. Thus, detecting their label errors can significantly increase their efficacy. We propose a novel framework, called CTRL (Clustering TRaining Losses for label error detection), to detect label errors in multi-class datasets. It detects label errors in two steps based on the observation that models learn clean and noisy labels in different ways. First, we train a neural network using the noisy training dataset and obtain the loss curve for each sample. Then, we apply clustering algorithms to the training losses to group samples into two categories: cleanly-labeled and noisily-labeled. After label error detection, we remove samples with noisy labels and retrain the model. Our experimental results demonstrate state-of-the-art error detection accuracy on both image (CIFAR-10 and CIFAR-100) and tabular datasets under simulated noise. We also use a theoretical analysis to provide insights into why CTRL performs so well.