We participated in the HIDA Datathon for grand challenges in climate change!
In the datathon, 94 data scientists from all over the world joined forces to advance five scientific projects on climate change using machine learning.
We focused on detecting anomalies in temperature and moisture measurements in the forest underground with data from the Hohes Holz forest in Sachsen-Anhalt (Spot the mistake in ~50 million data points, cleverly). Teaming up with Vilim Štih, we first tested a baseline solution using extended isolation forests and then moved on to modeling the sensor array data via 4D scalable Gaussian processes with the excellent library GPyTorch.
A huge shout-out to the amazing team at HIDA and the compute partners at SCC-KIT and Jülich SCC for the organization and support!