Classifying Long-tailed and Label-noise Data via Disentangling and Unlearning

In multimodal multimedia datasets, the challenges of long-tailed distributions and noisy labels often coexist, posing obstacles to model training and hindering performance. Existing studies on long-tailed noisy label learning (LTNLL) typically assume that the generation of noisy labels is independent of the long-tailed distribution, which may not be true from a practical perspective. However, the tail class samples are often observed to be mislabeled as head in real-world situations, exacerbating the original degree of imbalance. This phenomenon is termed “tail-to-head (T2H)” noise. T2H noise severely degrades model performance by polluting the head classes and forcing the model to learn the tail samples as head. To address this challenge, we investigate the dynamic misleading process of the noisy labels and propose a novel method called Disentangling and UNlearning for long-tailed noisy label lEarning (DUNE). It first employs the Inner-Feature Disentangling (IFD) to disentangle features internally. Based on this, Inner-Feature Partial Unlearning (IFPU) is then applied to weaken and unlearn incorrect feature regions correlated to wrong classes. This method prevents the model from being misled by noisy labels, enhancing the model's robustness against noise. To provide a controlled experimental environment, we further propose a new noise addition algorithm to simulate T2H noise. Extensive experiments on both simulated and real-world datasets demonstrate the effectiveness of our proposed method.