The Spotlight: A General Method for Discovering Systematic Errors in Deep Learning Models

d'Eon, Greg; d’Eon, Jason; Wright, James R.; Leyton‐Brown, Kevin

doi:10.1145/3531146.3533240

Cited by 29 publications

(12 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SliceFinder is one method that uses metadata to find slices with significantly high loss [15]. Often, there is not enough metadata to define slices with high error, so another family of methods uses model embeddings and clustering to find groups with high error [18,19]. Lastly, there are approaches that use end-user reports or crowd feedback to discover model failures or interesting behaviors [2,8,43].…”

Section: Model Evaluation Approachesmentioning

confidence: 99%

Zeno: An Interactive Framework for Behavioral Evaluation of Machine Learning

Cabrera,

Fu,

Bertucci

et al. 2023

Preprint

View full text Add to dashboard Cite

Figure 1: zeno is a framework for behavioral evaluation of machine learning (ML) models. It has two components, a Python API and an interactive UI. The API is used to generate information such as model outputs and metrics. Users then interact with the UI to see metrics, create slices, and write unit tests. In this toy example, a user is evaluating a cat and dog classifier. They see that the model has lower accuracy for dogs with pointy ears, and create a test expecting the slice accuracy to be higher than 70%.

show abstract

Section: Model Evaluation Approachesmentioning

confidence: 99%

Zeno: An Interactive Framework for Behavioral Evaluation of Machine Learning

Cabrera,

Fu,

Bertucci

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In addition to detecting spurious correlations, explainable AI techniques can also reveal human-subliminal signals about the disease and give us a better chance of addressing challenges posed by the pandemic. Lastly, slice discovery methods 67,68 , such as Domino 69 , can identify and describe semantically meaningful subsets of data on which the model performs poorly, thereby revealing spurious correlations.…”

Section: Model Building and Evaluationmentioning

confidence: 99%

Developing medical imaging AI for emerging infectious diseases

et al. 2022

View full text Add to dashboard Cite

Check for updatesAdvances in artificial intelligence (AI) and computer vision hold great promise for assisting medical staff, optimizing healthcare workflow, and improving patient outcomes. The COVID-19 pandemic, which caused unprecedented stress on healthcare systems around the world, presented what seems to be a perfect opportunity for AI to demonstrate its usefulness. However, of the several hundred medical imaging AI models developed for COVID-19, very few were fit for deployment in real-world settings, and some were potentially harmful. This review aims to examine the strengths and weaknesses of prior studies and provide recommendations for different stages of building useful AI models for medical imaging, among them: needfinding, dataset curation, model development and evaluation, and post-deployment considerations. In addition, this review summarizes the lessons learned to inform the scientific community about ways to create useful medical imaging AI in a future pandemic.nature communications (2022) 13:7060 | 1234567890():,; 1234567890():,;

show abstract

“…Prominent examples include facial recognition models failing to recognize women with dark skin tones [14] or translation models perpetuating gender stereotypes [118]. However, subgroups where a model fails can be highly contextual, specific, and may not match any social category (i.e., "men wearing thin framed glasses" [18] or "busy/cluttered workspace" [30]). It remains an open challenge for ML practitioners to detect which specific use case scenarios are likely to fail out of a possibly infinite space of model inputs -and prioritize which failures have the greatest potential for harm [7,43].…”

Section: Introductionmentioning

confidence: 99%

Angler: Helping Machine Translation Practitioners Prioritize Model Improvements

Robertson

Wang

Moritz

et al. 2023

Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

Angler enables ML developers to easily explore and curate challenge sets for machine translation. (A) The Table View lists all challenge sets, allowing users to compare them by metrics such as sample count, model performance, and familiarity score. After selecting a set, (B) the Detail View allows users to further explore samples in this set across various dimensions. (B1) The Timeline enables users to query data samples by time. (B2) The Spotlight presents visualizations with linking and brushing to help users characterize the set from different angles. (B3) The Sentence List shows all selected data samples and allows users to further fine-tune before exporting this challenge set for downstream tasks.

show abstract

The Spotlight: A General Method for Discovering Systematic Errors in Deep Learning Models

Cited by 29 publications

References 16 publications

Zeno: An Interactive Framework for Behavioral Evaluation of Machine Learning

Zeno: An Interactive Framework for Behavioral Evaluation of Machine Learning

Developing medical imaging AI for emerging infectious diseases

Angler: Helping Machine Translation Practitioners Prioritize Model Improvements

Contact Info

Product

Resources

About