Analysis of classifiers robust to noisy labels

TL;DR

This paper evaluates and compares robust classifiers designed to handle class-dependent label noise, demonstrating methods to estimate transition matrices and applying deep learning to real datasets to assess their effectiveness.

Contribution

It introduces an analysis of contemporary robust classifiers, including methods for estimating transition matrices, applied to deep learning on real datasets with unknown label noise.

Findings

Robust classifiers improve accuracy under noisy labels.

Transition matrix estimation enhances classifier performance.

Deep learning models show robustness on CIFAR with label noise.

Abstract

We explore contemporary robust classification algorithms for overcoming class-dependant labelling noise: Forward, Importance Re-weighting and T-revision. The classifiers are trained and evaluated on class-conditional random label noise data while the final test data is clean. We demonstrate methods for estimating the transition matrix in order to obtain better classifier performance when working with noisy data. We apply deep learning to three data-sets and derive an end-to-end analysis with unknown noise on the CIFAR data-set from scratch. The effectiveness and robustness of the classifiers are analysed, and we compare and contrast the results of each experiment are using top-1 accuracy as our criterion.