How Does Heterogeneous Label Noise Impact Generalization in Neural Nets?

TL;DR

This paper investigates how heterogeneous label noise impacts neural network generalization, revealing that noise affects only affected classes unless transfer occurs, filling a gap left by prior homogeneous noise studies.

Contribution

It introduces the first systematic study of heterogeneous label noise effects on neural networks across multiple datasets and tasks.

Findings

Label noise impacts only affected classes unless transfer occurs.

Heterogeneous noise effects differ from homogeneous noise.

Transfer between classes propagates label noise effects.

Abstract

Incorrectly labeled examples, or label noise, is common in real-world computer vision datasets. While the impact of label noise on learning in deep neural networks has been studied in prior work, these studies have exclusively focused on homogeneous label noise, i.e., the degree of label noise is the same across all categories. However, in the real-world, label noise is often heterogeneous, with some categories being affected to a greater extent than others. Here, we address this gap in the literature. We hypothesized that heterogeneous label noise would only affect the classes that had label noise unless there was transfer from those classes to the classes without label noise. To test this hypothesis, we designed a series of computer vision studies using MNIST, CIFAR-10, CIFAR-100, and MS-COCO where we imposed heterogeneous label noise during the training of multi-class, multi-task, and multi-label systems. Our results provide evidence in support of our hypothesis: label noise only affects the class affected by it unless there is transfer.