Open-set Label Noise Can Improve Robustness Against Inherent Label Noise

Abstract

Learning with noisy labels is a practically challenging problem in weakly supervised learning. In the existing literature, open-set noises are always considered to be poisonous for generalization, similar to closed-set noises. In this paper, we empirically show that open-set noisy labels can be non-toxic and even benefit the robustness against inherent noisy labels. Inspired by the observations, we propose a simple yet effective regularization by introducing Open-set samples with Dynamic Noisy Labels (ODNL) into training. With ODNL, the extra capacity of the neural network can be largely consumed in a way that does not interfere with learning patterns from clean data. Through the lens of SGD noise, we show that the noises induced by our method are random-direction, conflict-free and biased, which may help the model converge to a flat minimum with superior stability and enforce the model to produce conservative predictions on Out-of-Distribution instances. Extensive experimental results on benchmark datasets with various types of noisy labels demonstrate that the proposed method not only enhances the performance of many existing robust algorithms but also achieves significant improvement on Out-of-Distribution detection tasks even in the label noise setting.

Accepted by NeurIPS 2021

References in corpus (18)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Denoising Diffusion Probabilistic Models
• Temporal Ensembling for Semi-Supervised Learning
• Training Deep Neural Networks on Noisy Labels with Bootstrapping
• On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
• DivideMix: Learning with Noisy Labels as Semi-supervised Learning
• Deep Anomaly Detection with Outlier Exposure
• A Closer Look at Memorization in Deep Networks
• TurkerGaze: Crowdsourcing Saliency with Webcam based Eye Tracking
• Unsupervised Label Noise Modeling and Loss Correction
• Normalized Loss Functions for Deep Learning with Noisy Labels
• Robust Inference via Generative Classifiers for Handling Noisy Labels
• SELF: Learning to Filter Noisy Labels with Self-Ensembling
• Error-Bounded Correction of Noisy Labels
• Extended T: Learning with Mixed Closed-set and Open-set Noisy Labels
• Beyond Class-Conditional Assumption: A Primary Attempt to Combat Instance-Dependent Label Noise
• MetaInfoNet: Learning Task-Guided Information for Sample Reweighting