Resumen de Deep Teaching: Materials for Teaching Machine and Deep Learning
Christian Herta, Benjamin Voigt, Patrick Baumann, Klaus Strohmenger, Christoph Jansen, Oliver Fischer, Gefei Zhang, Peter Hufnagel
Machine learning (ML) is considered to be hard because it is relatively complicated in comparison to other topics of computer science. The reason is that machine learning is based heavily on mathematics and abstract concepts. This results in an entry barrier for students: Most students want to avoid such difficult topics in elective courses or self-study.
In the project Deep.Teaching we address these issues: We motivate by selected applications and support courses as well as self-study by giving practical exercises for different topics in machine learning. The teaching material, provided as jupyter notebooks, consists of theoretical and programming sections.
For didactical reasons, we designed programming exercises such that the students have to deeply understand the concepts and principles before they can start to implement a solution. We provide all necessary boilerplate code such that the students can primarily focus on the educational objectives of the exercises. We used different ways to give feedback for self-study: obscured solutions for mathematical results, software tests with assert statements, and graphical illustrations of sample solutions. All of the material is published under a permissive license. Developing jupyter notebooks collaboratively for educational purposes poses some problems. We address these issues and provide solutions/best practices.
