Toolbox for Uncertainty Quantification in Machine Learning - Master Practical (SS 2024)

The notion of uncertainty is of major importance in machine learning and constitutes a key element of modern machine learning methodology. In recent years, it has gained importance due to the increasing relevance of machine learning for practical applications, many of which are coming with safety requirements. In this regard, machine learning scholars have identified new problems and challenges, which call for new methodological developments. Indeed, while uncertainty has long been perceived as almost synonymous with standard probability and probabilistic predictions, recent research has gone beyond traditional approaches and also leverages more general formalisms and uncertainty calculi. For example, a distinction between different sources and types of uncertainty, such as aleatoric and epistemic uncertainty, turns out to be useful in many machine learning applications. The correct evaluation and analysis of methods from the field of uncertainty play a crucial role here, which, however, is not always straightforward to realize due to the lack of a standardized procedure to accomplish this. This is mainly due to subtleties of these methods as well as missing benchmark data sets for this field. The aim of the practical is to design and implement a development platform that enables the systematic evaluation and analysis of methods for the representation and quantification of uncertainties.

Prerequisites

• Software Engineering
• Introduction to Programming OR Programming and Modelling

General Information

Lecturer:

Eyke Hüllermeier

Language:

English

Time and place:

Lecture

Thursday

14-16

Scope:

6 SWH Master Practical

Target audience:

Master in Informatics

Assistant:

Timo Löhr