Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection
This work addresses label noise issues in machine learning, particularly for applications with noisy datasets, but it is incremental as it builds on existing meta-learning-based label correction methods.
The paper tackles the problem of label miscorrection and slow training in learning with noisy labels by proposing a method that learns a label transition matrix on the fly to mitigate miscorrection and uses a two-head architecture for efficient estimation within a single back-propagation, achieving the best training efficiency and comparable or better accuracy than existing methods.
Recent studies on learning with noisy labels have shown remarkable performance by exploiting a small clean dataset. In particular, model agnostic meta-learning-based label correction methods further improve performance by correcting noisy labels on the fly. However, there is no safeguard on the label miscorrection, resulting in unavoidable performance degradation. Moreover, every training step requires at least three back-propagations, significantly slowing down the training speed. To mitigate these issues, we propose a robust and efficient method that learns a label transition matrix on the fly. Employing the transition matrix makes the classifier skeptical about all the corrected samples, which alleviates the miscorrection issue. We also introduce a two-head architecture to efficiently estimate the label transition matrix every iteration within a single back-propagation, so that the estimated matrix closely follows the shifting noise distribution induced by label correction. Extensive experiments demonstrate that our approach shows the best performance in training efficiency while having comparable or better accuracy than existing methods.