Seminar: Mathematical Concepts of Machine Learning



+49 241 80-92064




  • Cycle: Summer semester
  • Assessment: written homework and presentation
  • Credits: 4 CP
  • Language: English

The use of machine learning and artificial intelligence has a huge potential for building future engineering systems such as autonomous robots, vehicles, or smart infrastructure systems. This is an interdisciplinary seminar offered in mechanical engineering and computer science due to the interdisciplinary nature of the topic. In addition to the challenges of designing ML and AI systems, it is critical to analyze the mathematical properties of these algorithms. This seminar will focus on some of the underlying mathematics of modern methods. You will learn about topics of current research in machine learning and focus on the underlying mathematical concepts. Many modern machine learning methods are readily implemented and can be used rather straightforwardly on a given data set. However, often this yields poor results. Thus, it is important to understand how the algorithms work and what mathematical properties of the data are critical to ensure good performance. A prominent example is correlated data, which can have dramatic consequences for certain methods and be manageable for others. The amount and quality of data is also an important aspect that heavily influences the success of the learning outcome.