Theory of Evidence in Active Learning

Dempster-Shafer theory of evidence allows to express both: alloteric uncertainty of a sample and epistemic uncertainty of the model. This information can become a base for efficient active learning algorithms. I will start with a demonstration of intuition behind the belief functions and basics of evidence theory. Then, I will discuss the state of the art algorithms for deep active learning related to the Dempster-Shafer theory. Finally, I will share my ideas for future research.

---------------------------------------

Plan wystąpień w tej edycji jest dostępny tutaj

The schedule of presentations can be checked here