In this episode of The Cool Data Projects Show, Neeraj Wagh from UIUC Bioengineering talks about his work on evaluating the robustness of ML models which was recently published at NeurIPS 2022. His project shows how domain expertise can be exploited to assess model robustness without access to any external data or expensive expert labels. Although he focuses on brain EEG signals, his approach is general and can be easily adapted to other machine learning application domains. Neeraj shares the broader motivation and inspiration for the work, and highlights the 3 key steps of the proposed robustness evaluation approach.
The Cool Data Projects Show is hosted by Comet ML Developer Advocate Kristen Kehrer. Learn more at www.comet.com .
Ещё видео!