Using Early-Learning Regularization to Classify Real-World Noisy Data

TL;DR

This paper investigates the effectiveness of Early-Learning Regularization in classifying noisy real-world data, confirming its robustness and exploring enhancements with Sharpness-Aware Minimization, aiming to improve accuracy in noisy datasets.

Contribution

It replicates and validates Early-Learning Regularization on real-world noisy data and combines it with Sharpness-Aware Minimization for improved performance.

Findings

Consistent experimental results with original claims

14.6% accuracy improvement with Sharpness-Aware Minimization

Potential for further improvements with larger datasets and cleaner data

Abstract

The memorization problem is well-known in the field of computer vision. Liu et al. propose a technique called Early-Learning Regularization, which improves accuracy on the CIFAR datasets when label noise is present. This project replicates their experiments and investigates the performance on a real-world dataset with intrinsic noise. Results show that their experimental results are consistent. We also explore Sharpness-Aware Minimization in addition to SGD and observed a further 14.6 percentage points improvement. Future work includes using all 6 million images and manually clean a fraction of the images to fine-tune a transfer learning model. Last but not the least, having access to clean data for testing would also improve the measurement of accuracy.