Computational aspects of machine learning

Course goals

After completion of the course, the student is expected to understand:
1. the statistical and mathematical methods at the foundation of machine learning
techniques
• Bayesian theory
• Linear regression, logistic regression and classification
• Clustering and dimensionality reduction
• Ensemble method
• Kernel methods
• Neural network and deep learning
2. How to choose the best learning strategy
• Supervised, unsupervised or reinforced?
• Which learning algorithm?
3. How to apply machine learning to a physics project
• The applications of machine learning in different fields of experimental physics
• How to execute a practical project
• How to obtain, interprete and report the results of the project in a complete and
scientifically correct way

Content

The masters’ course “Computational aspects of machine learning” aims to familiarize the
physics students with machine learning algorithms, the mathematical and statistical methods
at their foundation and with their the applications to experimental physics research. In the first
part of the course (~60%) we will cover the basic theory and statistical methods at the
foundation of the most used learning algorithms. The depth of the theory part will be calibrated
to the needs of a typical master student in physics. During the last 40% of the course, we will
switch from theory to practical work. The students will be assigned to real projects that require
a basic knowledge of either python or C++. During the projects period, guest lectures from
researchers in different fields of experimental physics will highlight real applications to cuttingedge
research.
Students will be assigned to the projects in small groups. Each group is supposed to develop
the respective project up to a satisfactory level. The assignment will happen after the midterm
evaluation. The students will be guided through the project, receive feedback on the
progresses and technical help during the tutorial sessions. Finally, the students’
performances and learning progresses will be evaluated by means of mid-term evaluation for
the theory course part one and final project report for the project-oriented course part two.