Machine learning and artificial intelligence in applied and fundamental sciences – practical methods, applications and critiques in a nutshell

Speaker: Alexander Struck – Rhein-Waal University

Host: Dirk Schuricht

Abstract:

The use of statistical and numerical algorithms that scrutinise large data set in order to extract common properties of datapoints, classify and quantify underlying structures and data-generating processes — widely known as machine learning — is nowadays readily and freely available in highly advanced software systems that run on all scales of hardware. Very often it is used in more or less empirical approaches to evaluate heterogeneous and sometimes incomplete numerical and categorical evidence resulting from experimental data or numerical simulations. Models obtained from these treatments are used to predict new outcomes and to describe the nature of the data.

In this talk, I will give a rough overview about common methods that can be employed in the context of natural sciences and point out a successful application of machine learning for detection of bacterial contamination using Raman spectroscopy that has been developed recently in cooperation with the food-processing industry. I will also raise the question whether machine learning fits into the well-established reasoning procedure and how it might alter the way we need to think about data and data generation.