May 6, 2024Open Access

Annot-Mix: Learning with Noisy Class Labels from Multiple Annotators via a Mixup Extension

View Full Paper

Ask AI

Bookmark

Authors

MHMarek HerdeUniversity of Kassel LLLukas LührsUniversity of Kassel DHDenis HuseljicUniversity of Kassel

Discussion

Member takes

Implication

Benchmark analysis reveals that annot-mix improves neural network robustness against noisy class labels across eleven datasets, highlighting superior multi-annotator learning.

Key Points

Annot-mix framework outperforms eight state-of-the-art methods across eleven datasets containing noisy class labels from human crowdworkers and simulated annotator sources.
Evaluation using a multi-annotator mixup extension incorporates annotator identity directly into regularization during neural network training to mitigate false label memorization.
Integrating annotator-aware data augmentation into deep learning pipelines may enable more reliable classification when crowdsourced training data contains label noise.

Cite This Study

Herde et al. (2024) studied this question.

synapsesocial.com/papers/68e6b6eab6db64358763871d https://doi.org/10.48550/arxiv.2405.03386

View Full Paper

Ask AI

Bookmark

Implication

Benchmark analysis reveals that annot-mix improves neural network robustness against noisy class labels across eleven datasets, highlighting superior multi-annotator learning.

Annot-Mix: Learning with Noisy Class Labels from Multiple Annotators via a Mixup Extension

Discussion

Member takes

Implication

Key Points

Cite This Study

Annot-Mix: Learning with Noisy Class Labels from Multiple Annotators via a Mixup Extension

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider