Scientific machine learning for data-driven discovery

We give practical workshops to explain how science can benefit from machine learning — both at an introductory level and on particularly relevant ML techniques.

• Introduction to Machine Learning
• Technical ML Workshop

If you are interested in participating in one of our workshops, send us an email.

We advise and collaborate with domain researchers. Here are some of our latest projects

• PollenClim: Working with archaeologists and paleoclimatologists, we improve our understanding of the past climate from a combination of simulations and fossilized pollen data

Techniques: Gaussian Processes, AutoML

• CoBaRd: We are currently working with physicists to design a device (moderator) to reduce cosmogenic background by finding the parameter set that is predicted to maximally reduce the number of neutrons as predicted by Monte Carlo simulations in GEANT.
• LiTrace: We are currently working physicists to obtain position (xyz), time and energy, parameters of scintillation from the currents generated at photodetectors (NNVT MCP-PMTs) placed on the wall of the vessel.

We develop free software for machine learning, either as contributions to existing projects or as new projects to address ML software gaps for scientific work.

• InferenceObjects.jl
• Pathfinder.jl
• Blogposts

The Machine Learning ⇌ Science Colaboratory is part of the Cluster of Excellence Machine Learning: New Perspectives in Science at the University of Tübingen. We are a passionate team of researchers and engineers working to increase the impact of machine learning (ML) on the sciences and the humanities.