The talk starts at 12:15.

Please note that due to COVID-19, only the first 20 NR employees will be able to see this talk in person. The rest of the participants can watch the streamed talk on Zoom with a link (link to come).

For NR employees that wish to attend the talk in person, please respond here: https://docs.google.com/forms/d/e/1FAIpQLSfA0AtVr9xs7G8qOzmzY4e8QHK_OvPOvlEheEUJk0OAAGKUWA/viewform?usp=sf_link

Speaker: Kristoffer Knutsen Wickstrøm (UiT)

Location: NR (Spiseriet)

ZOOM: link: https://uio.zoom.us/j/69618440273

Meeting ID: 696 1844 0273

Title: Uncertainty and interpretability in deep learning-based support systems for medical data

Abstract: 

Deep learning-based support systems have in recent years demonstrated impressive performance on numerous clinical tasks involving both image and time series data. However, they lack crucial components that are necessary for developing dependable and trustworthy systems, namely uncertainty and interpretability. This presentation will illustrate why uncertainty and interpretability is important when applying a deep learning approach to a given task, and particularly for healthcare tasks. Furthermore, recent research on uncertainty and interpretability in deep learning for both image and time series will be presented, where uncertainty in interpretability will be the main focus.

(Note that Kristoffer will not be presenting in person but through Zoom).