Learning from Long-Tailed Noisy Data with Sample Selection and Balanced Loss

no code implementations • 20 Nov 2022 • Lefan Zhang, Zhang-Hao Tian, Wujun Zhou, Wei Wang

Specifically, we separate the noisy training data into clean labeled set and unlabeled set with sample selection, and train the deep neural network in a semi-supervised manner with a balanced loss based on model bias.

Towards Understanding Deep Learning from Noisy Labels with Small-Loss Criterion

no code implementations • 17 Jun 2021 • Xian-Jin Gui, Wei Wang, Zhang-Hao Tian

However, there are few theoretical analyses to explain why these methods could learn well from noisy labels.