Workshop on Machine Learning and Neural Networks (Sep. 2018)
In recent years computational techniques, such as Neural Networks and other Machine learning methods have become increasingly fashionable also for meteorological applications. Due to the high interest in the topic and to exploit its potential, the ECS felt the need to be further trained in these methods. So they organized a workshop on Machine Learning and Neural Networks from September 24th - 26th 2018 at the Schneefernerhaus on the Zugspitze. The 22 participants and 3 instructors found perfect study conditions in the calm mountain atmosphere.
Together with the invited guest Martin Felder (Zentrum für Sonnenenergie und Wasserstoff-Forschung Baden-Württemberg), Stephan Rasp (W2W) and Sebastian Lerch (W2W) designed a very dense and highly instructive program. A mixture of basic and advanced lectures and well explained hands-on exercises lead to a stimulating environment. Starting with basic linear regression models, the ECS learned to apply highly complex techniques such as random forests, embedded neural networks, and convolutional neural networks with multiple hidden layers using the python libraries “keras” and “scikit_learn”. At the end of each of the 6 hands-on sessions, the ECS had sufficient time to apply their own ideas based on the techniques learned. To increase the motivation, the ECS evaluated their results in a competition using a web application, where they fought for the first places, i.e. the best trained models.
Furthermore, the ECS learned about the research performed at the Schneefernerhaus during a guided tour, they enjoyed lunch, sunny weather, and panoramic views on the Zugspitz-summit, and finally went sledding on Zusgspitzplatt together.