Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection

Seong Min Kye, Kwanghee Choi, Joonyoung Yi, Buru Chang

Keywords: Noisy Labels, Label Correction, Meta-Learning

Abstract: 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 meta-learning 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 learn the label transition matrix every iteration within a single back-propagation, so that the matrix estimate closely follows the shifting 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.

One-sentence Summary: A robust and efficient meta-learning method that learns a label transition matrix on the fly