Machine learning (ML) uses patterns hidden in data to make predictions that rival or sometimes even surpass human performance. What are these patterns? Popular ML-methods, such as deep learning, do not provide any reasonable answer to this question; they speak a language that is impossible to understand for us humans. What if we would require them to speak understandably? What if we define an algorithm that explicitly names the patterns it sees in the data, such that we humans can explore what is going on in our data? In this talk I will discuss an algorithm, called subgroup discovery, that does exactly this. Besides introducing it in layperson terms, I will give examples on how have we used it to gain insights in topics like voting behavior, disaster survival, medical treatment effectiveness, and (materials) science.
Zoom-Link zu dieser Session:https://zoom.us/j/95377268993Bitte registrieren Sie sich auf der SciCAR-Homepage, um an der virtuellen Konferenz teilnehmen zu können. Das benötigte Passwort zum Zoom-Meeting erhalten angemeldete Teilnehmer am Montagmorgen (02.11.) per E-Mail. Eine Anmeldung ist auch noch nach Konferenzbeginn möglich.