Validity in Machine Learning: A Software Engineering Perspective
Machine Learning and AI is becoming ubiquitous today. More and more decisions that affect only daily lives are made by machines and algorithms. It is therefore important that these decisions are made based on a valid and fair model of the world.
In this talk, we are taking a software engineering’s perspective of building a Machine-Learned model and discuss several choices we have to make during the process of model creation that all threaten the validity of the model’s predictions. The goal of this talk is to raise awareness of validity in Machine Learning and to foster further research in this area.
Norbert Siegmund holds the Chair of Software Systems at Leipzig University. Prof. Siegmund received his PhD with distinction in 2012 from the Otto-von-Guericke University Magdeburg. His research aims at the automation of software engineering by combining methods from software analysis, Machine Learning, and meta-heuristic optimization. His special interests include software product lines and configurable software systems, performance and energy optimization, and digitization based on MicroServices and Web technologies. He is author and co-author of more than 70 peer-reviewed scientific publications. He regularly serves in program committees of top-ranked international conferences and is in the review board of the renowned journal IEEE Transactions on Software Engineering. He is a founding member of the Java User Group Thuringia.
Back to the Summer School 2020 overview