Identifying Mislabeled Instances in Classification Datasets

TL;DR

This paper introduces a non-parametric pipeline for detecting mislabeled instances across various data types, significantly improving data quality for supervised learning.

Contribution

It presents a novel end-to-end method for identifying mislabeled data in multiple modalities, with quantitative evaluation and real-world dataset application.

Findings

Achieves over 0.84 average precision in identifying mislabeled instances

Successfully finds mislabeled data in CIFAR-100 and Fashion-MNIST

Provides publicly available code and implementation

Abstract

A key requirement for supervised machine learning is labeled training data, which is created by annotating unlabeled data with the appropriate class. Because this process can in many cases not be done by machines, labeling needs to be performed by human domain experts. This process tends to be expensive both in time and money, and is prone to errors. Additionally, reviewing an entire labeled dataset manually is often prohibitively costly, so many real world datasets contain mislabeled instances. To address this issue, we present in this paper a non-parametric end-to-end pipeline to find mislabeled instances in numerical, image and natural language datasets. We evaluate our system quantitatively by adding a small number of label noise to 29 datasets, and show that we find mislabeled instances with an average precision of more than 0.84 when reviewing our system's top 1\% recommendation. We then apply our system to publicly available datasets and find mislabeled instances in CIFAR-100, Fashion-MNIST, and others. Finally, we publish the code and an applicable implementation of our approach.