What question did this study set out to answer?

The aim is to develop a training method for deep learning models that mitigates the impact of noisy labels.

January 1, 2018

Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial examines co-teaching to enhance robustness against noisy labels in deep learning.

The aim is to develop a training method for deep learning models that mitigates the impact of noisy labels.
Introduced Co-teaching paradigm involving two deep neural networks training simultaneously.
Each network selects possibly clean labels from data in mini-batches and communicates them.
Backpropagation updates occur based on selected data from peer networks.
Co-teaching method shows significant improvement in model robustness over state-of-the-art approaches.
Empirical tests on datasets MNIST, CIFAR-10, and CIFAR-100 validate effectiveness.

Han et al. (2018) studied this question.

View Full Paper

Ask AI

Bookmark