Tightening the Loop in Mixed-Initiative ML Engineering and Domain Annotation using Active Learning and Visual Analytics



 View presentation:2022-10-17T14:55:00ZGMT-0600Change your timezone on the schedule page
2022-10-17T14:55:00Z

Exemplar figure, described by caption below — This figure illustrates the dashboard workflow. The data annotation loop involves domain experts who annotate the data through frontend and receive feedback after the backend machine learning (ML) engine processes, which data are most important to annotate next. The model verification loop involves ML experts, who configure model parameters based on the already-labelled data. The two loops are tightened by constant feedback from one another: domain experts make more annotated data available and learn which further data is needed, while ML experts learn about how the model is currently performing based on the available data and configure the model accordingly.

The live footage of the talk, including the Q&A, can be viewed on the session page, BioMedical AI: Keynote and Session 1.

Abstract