|
Instructors: Dr. Kristof Schmieden
Event type:
Lecture
Displayed in timetable as:
08.128.614
Hours per week:
2
Credits:
3,0
Language of instruction:
Englisch
Min. | Max. participants:
- | -
Requirements / organisational issues:
No pre-requirements are needed to attend the course.
As all examples are implemented using python, running as jupyter notebook or on Googles CoLab, basic knowlegde of python is highly recommended.
Contents:
The fundamentals of mashine learning will be discussed briefly at the beginning of the course.
The focus will be put on the practical implementation and usage of ML tools for classification and regression as well as the usage of genrative algorithms. To this end we will implement many examples using Python & Keras and explore their limits. Example data used will be data from particle physics besides the well known ML training data. However, no detailed knowledge of the physics processes is needed to follow the lecture.
During the course of the lecture you will learn, amongst other:
- When a ML classificator is useful
- How a ML tool is implemented and used (Keras neural networks)
- How to see if a training is robust or not
- How to estimate uncertainties on ML methods
- How to optimise ML classifications / regressions
- Using Generative algorithms in event simulation
- Adversarial Learning (GANs)
At the End of the lecture there will be time to discuss one more topic of your choice, like explorative algorithms, boosted decision trees, or other.
Recommended reading list:
Book suggestion: Martin Erdmann, Deep learning for physics research
https://hds.hebis.de/ubmz/Record/HEB487091906
|
