Sharon Fogel- Are you sure? Estimating uncertainty in classification networks

Collaboration with women in AI -

Join Zoom Meeting
https://zoom.us/j/6043600514?pwd=VTFuU2VSTTNhTE1RRFJTZjhZNTN1Zz09

Meeting ID: 604 360 0514
Password: 703769

Recently, Deep Neural Networks have achieved tremendous improvement in tasks such as image classification and text recognition.
Nonetheless, one remaining problem is estimating how confident the model is.
Often, DNNs tend to be overconfident about their predictions, which in might be problematic for real world applications, such as medical imaging or autonomous driving.
Knowing when your model is uncertain will allow adding human intervention in such cases.
In this talk, we will go over different approaches to measure the uncertainty estimation in your model and improve it, focusing on classification networks.