Model-Agnostic Label Quality Scoring to Detect Real-World Label Errors (Q7433313)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Unfortunately, we do not yet have an article page for this item.

Resource published at Zenodo repository.
Language Label Description Also known as
default for all languages
No label defined
    English
    Model-Agnostic Label Quality Scoring to Detect Real-World Label Errors
    Resource published at Zenodo repository.

      Statements

      We consider algorithms to find wrongly labeleddata, which lurks in many real-world applicationsand hampers training/evaluation of ML models. We present the first empirical study of various scoring methods for this task on real datasets with naturally-occurring label errors (as opposed tosynthetically introduced label errors). The label quality scores considered here can be utilized with arbitrary classification models. We examine fivepopular image recognition models (and ensembles thereof) to comprehensively characterize how well different scores detect label errors in practice.
      0 references
      Johnson Kuan
      0 references
      Jonas Mueller
      0 references

      Identifiers