Multi-Objective Interpolation Training for Robustness to Label Noise

Deep neural networks trained with standard cross-entropy loss memorize noisy labels, which degrades their performance. Most research to mitigate this memorization proposes new robust classification loss functions. Conversely, we propose a Multi-Objective Interpolation Training (MOIT) approach that jointly exploits contrastive learning and classification to mutually help each other and boost performance against label noise. We show that standard supervised contrastive learning degrades in the presence of label noise and propose an interpolation training strategy to mitigate this behavior. We further propose a novel label noise detection method that exploits the robust feature representations learned via contrastive learning to estimate per-sample soft-labels whose disagreements with the original labels accurately identify noisy samples. This detection allows treating noisy samples as unlabeled and training a classifier in a semi-supervised manner to prevent noise memorization and improve representation learning. We further propose MOIT+, a refinement of MOIT by fine-tuning on detected clean samples. Hyperparameter and ablation studies verify the key components of our method. Experiments on synthetic and real-world noise benchmarks demonstrate that MOIT/MOIT+ achieves state-of-the-art results. Code is available at https://git.io/JI40X.

1. Introduction

Building a new dataset usually involves manually labeling every sample for the particular task at hand. This process is cumbersome and limits the creation of large datasets, which are usually necessary for training deep neural networks (DNNs) in order to achieve the required performance. Conversely, automatic data annotation based on web search and user tags [29], [22] leverages the use of larger data collections at the expense of introducing some incorrect labels. This label noise degrades DNN performance [3], [52] and this poses an interesting challenge that has recently gained a lot of interest in the research community [45], [41], [23], [50], [12], [1], [28], [55], [13], [31].