Skip to content

Read-only version: For security reasons, the Research Information System (FIS) has been temporarily replaced by this read-only version. All public content and full texts remain available; login, editing, submission and statistics are not available. Questions: fis@uni-bamberg.de

Home › Publications › An Embedding is Worth a Thousand Noisy Labels

Publication

An Embedding is Worth a Thousand Noisy Labels

Di Salvo, Francesco; Doerrich, Sebastian; Rieger, Ines; u. a. (2025): An Embedding is Worth a Thousand Noisy Labels, in: Transactions on machine learning research, Amherst, Massachusetts: OpenReview.net, Nr. 3, S. 1–19.

Details

Faculty/Chair
Faculty/Professorship
Author
By
... ; Doerrich, Sebastian; ...
Title of the Journal
Transactions on machine learning research
ISSN
2835-8856
Publisher Information
OpenReview.net : Amherst, Massachusetts
Year of publication
2025
Issue
3
Language
English
URL
Abstract

The performance of deep neural networks scales with dataset size and label quality, rendering the efficient mitigation of low-quality data annotations crucial for building robust and cost-effective systems. Existing strategies to address label noise exhibit severe limitations due to computational complexity and application dependency. In this work, we propose WANN, a Weighted Adaptive Nearest Neighbor approach that builds on self-supervised feature representations obtained from foundation models. To guide the weighted voting scheme, we introduce a reliability score, which measures the likelihood of a data label being correct. WANN outperforms reference methods, including a linear layer trained with robust loss functions, on diverse datasets of varying size and under various noise types and severities. WANN also exhibits superior generalization on imbalanced data compared to both Adaptive-NNs (ANN) and fixed k-NNs. Furthermore, the proposed weighting scheme enhances supervised dimensionality reduction under noisy labels. This yields a significant boost in classification performance with 10x and 100x smaller image embeddings, minimizing latency and storage requirements. Our approach, emphasizing efficiency and explainability, emerges as a simple, robust solution to overcome inherent limitations of deep neural network training.

GND Keywords
Nächste-Nachbar-Verfahren
Bildanalyse
Medizin
Keywords
Adaptive kNN
Label noise
Robustness
Foundation models
DDC Classification
004 Computer science
RVK Classification
ST 300
Peer Reviewed
yes
International Distribution
yes
Open Access Journal
yes
Type
Article
URI
Activation date
2025-08-06

Permalink: https://fis.uni-bamberg.de/handle/uniba/109535 · Data as JSON · Cite: BibTeX · RIS

Data as of: 2026-11-15

Restricted read-only access: login, editing and submission are temporarily disabled.