ProSelfLC: Progressive Self Label Correction Towards A Low-Temperature Entropy State
Xinshao Wang, Hua Yang, Elyor Kodirov, Sankha Subhra Mukherjee, David Andrew Clifton, Neil M. Robertson
ProSelfLC improves DNN robustness by progressively trusting low-entropy predictions and correcting labels during training.
bioRxiv · Jul 1, 2022 · PreprintBioinformatics