Institute of Solid State Physics


SS22WS22SS23WS23SS24WS24      Guidelines for Master Students

Feature Selection in Materials Design
Johannes Cartus
Institute of Solid State Physics
11:15 - 12:15 Wednesday 04 March 2020 PH 01 150

A first step that needs to be taken when deriving a model is to assess which properties of a physical system are coupled and how they influence each other. The workflow of natural scientists is thus usually to make observations, look for correlations in these observations, setup hypotheses to explain the correlations and apply Occam’s razor. The result is a model that can be used to predict new, testable observations to validate or invalidate the hypothesis. Statistics and big data technologies allow to formalize the search for correlations and model building if enough data is available.

Feature Selection is a branch of data science which focuses on finding the properties that affect a given target property (so-called features), while Feature Extraction aims to find the functional relationships between these affecting properties and the target. In this talk I will give an overview over recent developments in both fields and introduce a few applications (with a focus on ab-initio surface science and my own work).