Friends and Foes in Learning from Noisy Labels

TL;DR

This paper critiques current evaluation practices for learning from noisy labels, proposes new datasets and metrics, and introduces an improved method that significantly enhances performance on both synthetic and real-world noisy datasets.

Contribution

It introduces valid evaluation metrics and datasets for noisy label learning, and develops a new method combining beneficial components to improve accuracy.

Findings

F&F method outperforms existing approaches on nCIFAR datasets

Proposed metrics and datasets better evaluate noisy label learning

Significant accuracy improvements on Clothing1M dataset

Abstract

Learning from examples with noisy labels has attracted increasing attention recently. But, this paper will show that the commonly used CIFAR-based datasets and the accuracy evaluation metric used in the literature are both inappropriate in this context. An alternative valid evaluation metric and new datasets are proposed in this paper to promote proper research and evaluation in this area. Then, friends and foes are identified from existing methods as technical components that are either beneficial or detrimental to deep learning from noisy labeled examples, respectively, and this paper improves and combines technical components from the friends category, including self-supervised learning, new warmup strategy, instance filtering and label correction. The resulting F&F method significantly outperforms existing methods on the proposed nCIFAR datasets and the real-world Clothing1M dataset.