Coresets for Robust Training of Neural Networks against Noisy Labels

Abstract

Modern neural networks have the capacity to overfit noisy labels frequently found in real-world datasets. Although great progress has been made, existing techniques are limited in providing theoretical guarantees for the performance of the neural networks trained with noisy labels. Here we propose a novel approach with strong theoretical guarantees for robust training of deep networks trained with noisy labels. The key idea behind our method is to select weighted subsets (coresets) of clean data points that provide an approximately low-rank Jacobian matrix. We then prove that gradient descent applied to the subsets do not overfit the noisy labels. Our extensive experiments corroborate our theory and demonstrate that deep networks trained on our subsets achieve a significantly superior performance compared to state-of-the art, e.g., 6% increase in accuracy on CIFAR-10 with 80% noisy labels, and 7% increase in accuracy on mini Webvision.

References in corpus (13)

• mixup: Beyond Empirical Risk Minimization
• Wide Residual Networks
• Understanding deep learning requires rethinking generalization
• Training Deep Neural Networks on Noisy Labels with Bootstrapping
• WebVision Database: Visual Learning and Understanding from Web Data
• Early stopping and non-parametric regression: An optimal data-dependent stopping rule
• Gradient Descent with Early Stopping is Provably Robust to Label Noise for Overparameterized Neural Networks
• Understanding and Utilizing Deep Neural Networks Trained with Noisy Labels
• Generalization Error Bounds of Gradient Descent for Learning Over-parameterized Deep ReLU Networks
• IMAE for Noise-Robust Learning: Mean Absolute Error Does Not Treat Examples Equally and Gradient Magnitude's Variance Matters
• Greedy Column Subset Selection: New Bounds and Distributed Algorithms
• Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization
• Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee

Cited by in corpus (5)

• Sample Selection with Uncertainty of Losses for Learning with Noisy Labels
• Deep Learning on a Data Diet: Finding Important Examples Early in Training
• Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization
• GRAD-MATCH: Gradient Matching based Data Subset Selection for Efficient Deep Model Training
• Learning to Combat Noisy Labels via Classification Margins