Communications of the ACM
ACM Opinion
Quantifying Uncertainty in Deep Learning
Because deep learning lacks the notion of uncertainly, it doesn't necessarily perform well when it encounters unfamiliar situations. It might predict outcomes with high confidence even though it doesn't have any experience on which to base those predictions.
In a video, Aalto University Professor Arno Solin and doctoral student Lassi Meronen discuss developing better deep learning models and how to account for uncertainty.
From Aalto University
View the Video
Comments
Amir Mukeri
Sounds interesting. Any recommendation on recent papers on this topic would be helpful.
Displaying 1 comment