Speaker
Patrick Bos
Description
With the boom of applications of deep neural networks in (astronomical) research, we have grown an urgent need to crack open our own black boxes, lest our reviewers start asking us difficult questions we cannot answer. Enter the field of "explainable AI" (XAI), in which methods to solve this issue are developed and studied. I will give a short introduction into XAI in general, and present our specific method called Distance Explainer, which "explains" which parts of your data actually drive (dis)similarity in a learned embedding space. I'm curious whether you are working on something where this could be useful. If so, let's talk.
| Talk category | Splinter 6: Data science |
|---|
