Maschinelles Lernen auf der atomaren Skala

What it is about

You learn how machine learning is applied to atomic systems (molecules, solids): from data generation with DFT/quantum chemistry through suitable atomistic descriptors to machine-learned interatomic potentials and Hamiltonian learning. In the end you will be able to analyze datasets, select appropriate representations, train and evaluate ML models, and carry out your own atomic ML project.

What you will be able to do

• Learn the significance of ML for atomic systems
• Understand atomistic data generation
• Gain an overview of atomistic descriptors
• Gain experience with machine-learned interatomic potentials
• Get to know various ML applications for atomic systems
• Gain practical experience with Python notebooks
• Apply different ML methods in Python
• Carry out fundamental data analyses
• Tackle atomic problems with suitable ML approaches
• Be able to follow presentations on atomic ML
• Identify atomic research questions for ML
• Select appropriate representations as input
• Justify method choice for representations
• Evaluate and improve the performance of ML models
• Carry out an atomistic ML project
• Critically comment on ML applications

What the module consists of

• Vorlesungconveys foundations and concepts of atomic machine learning
• Computerübungen / Tutorials (Python-Notebooks)illustrate practical examples and convey application
• Projektarbeitenables independent work on an atomic ML problem

Teaching method

• Vorlesungenintroduce theory and methods
• Computerübungenpractical application of methods in Python notebooks
• Projektarbeitdeepened, independent application and transfer to research questions