Rethinking the Value of Labels for Instance-Dependent Label Noise Learning
This addresses the problem of noisy labels degrading deep learning performance for practitioners dealing with real-world data where noise depends on both features and true labels.
The paper tackles instance-dependent label noise in datasets by proposing a deep generative model that avoids explicit noise transition matrix modeling, achieving significant performance improvements over state-of-the-art methods on synthetic and real-world datasets.
Label noise widely exists in large-scale datasets and significantly degenerates the performances of deep learning algorithms. Due to the non-identifiability of the instance-dependent noise transition matrix, most existing algorithms address the problem by assuming the noisy label generation process to be independent of the instance features. Unfortunately, noisy labels in real-world applications often depend on both the true label and the features. In this work, we tackle instance-dependent label noise with a novel deep generative model that avoids explicitly modeling the noise transition matrix. Our algorithm leverages casual representation learning and simultaneously identifies the high-level content and style latent factors from the data. By exploiting the supervision information of noisy labels with structural causal models, our empirical evaluations on a wide range of synthetic and real-world instance-dependent label noise datasets demonstrate that the proposed algorithm significantly outperforms the state-of-the-art counterparts.