Prof. Ulrike von Luxburg

ZOOM Meeting

January 14th, 2021

2 pm

Chair: Prof. Matthias Rarey

The Inductive Bias in Machine Learning

All machine learning algorithms necessarily have an inductive bias. While this might not be a problem if all we want to do is predict some customer’s behavior (as long as our predictions work well), in science we want to discover ``ground truth’’ and to ``understand’’ the underlying phenomena that generate the data. In particular in applications where data is scarce, as it is the case in many fields of science, the inductive bias might have a considerable influence on the outcome of machine learning algorithms. However, in many fields of machine learning it is not even clear how we could specify, understand or evaluate the inductive bias of an algorithm, let alone understand what the bias is. This is what I am going to discuss in my talk.