ML3 seminar
Decomposing
Neural   Networks
An applicant's guide to artificial learning.
Decomposing Neural Networks
is a course of the Leuphana University of Lüneburg for the Bachelor program of Digital Media.
Content: Neural networks can generate artworks or beat the world champion in go by suggesting moves that no expert would consider. Metaphorical they are often described as black boxes, which are generally seen as uninterpretable by humans. But how black is the black box really? Why are those models so powerful? And where are their limits or weaknesses? Together we will look at the inner mechanism of neural networks, but likewise, dive into where the concepts are derived from. We will practically experiment with different architectures to deepen the understanding of the concepts, exploring their power but also their limitations. see Sessions
Learning Objectives: This course aims to familiarize students with the technological background of machine learning in an intuitive way. Students of this course should learn to:
Assignments and workload: This course has a workload of 5 ETC. It comprises pre- and post-lecture assignments during the duration of the course, which will encompass reading assignments as well as experimental coding tasks.
Exam: Students have to train an own neural network (based on the code developed together in the seminar) and critically reflecting it in written form.
Students about this seminar: