PhD Colloquium - Abstracts

Talk

Monday 26th of June 2023, 13:30-14:30 in HG02.052
Laurens Sluijterman
How to trust a neural network?

Abstract

In the thrilling world of artificial intelligence, nothing has been quite as transformative as the rise of neural networks. This talk, ‘How to trust a neural network?', journeys into the heart of these digital marvels. We start by laying the groundwork with the basics of neural networks, breaking down their complex mechanisms into digestible nuggets. Armed with this understanding, we then step onto the exciting terrain of uncertainty estimation. Here, we'll explore an array of techniques used to measure model uncertainty, providing you with a holistic understanding of how to trust these AI powerhouses. The highlight? We'll unpack our new method that uses a likelihood ratio approach.
This abstract was written by GPT4.


(Back to PhD Colloquium schedule)