Introduction to Neuronal Networks for Physicists
PH0101 is a semester module in German language at Bachelor’s level which is offered in winter semester.
If not stated otherwise for export to a non-physics program the student workload is given in the following table.
Responsible coordinator of the module PH0101 is the Dean of Studies at Physics Department.
Content, Learning Outcome and Preconditions
- Fundamentals of scientific programming
- data structures
- LU decomposition of matrices
- algorithms for PDEs
- Neural Networks
- basics: from perceptrons to sigmoid neurons
- example: classification of points in a plane with 4 neurons
- back propagation / training of neural networks
- example: handwritten digit recognition (MNIST)
- extentions: convolutional neural networks etc.
- project work
- choosing a project (see Moodle)
- working on the project under the supervision of the tutors
- presentation of the results
After successful completion of the module the students are able to:
- understand and use basic numerical algorithms
- understand the workings for neural networks
- assess whether a given problems is suitable for study with neural networks
- implement simple neural networks from scratch
- implement complex neural networks with the help of libraries
Basic programming skills are recommended.
Courses, Learning and Teaching Methods and Literature
Courses and Schedule
|Introduction to Neuronal Networks for Physicists
Wed, 16:00–18:00, PH HS3
|Exercise to Introduction to Neuronal Networks for Physicists
Responsible/Coordination: Recksiegel, S.
|dates in groups
Learning and Teaching Methods
In the lecture the content is first developed on an electronic whiteboard (the slides can be downloaded as PDF after each lecture) and then presented as Python/Mathematica code.
Exercise sheets often involve the reproduction of algorithms developed in the lecture, they are first studied individually and then discussed in groups.
In the last third of the term the students work on a project of their choice (with help from the tutors) and finally present their results to the class.
Electronic Whiteboard, demonstrations in Mathematica, Python and Keras/Tensorflow;
Exercise sheets. Web site: https://www.moodle.tum.de/course/view.php?id=79979
- Michael A. Nielsen, "Neural Networks and Deep Learning", Determination Press 2015, http://neuralnetworksanddeeplearning.com/
- David MacKay, "Information Theory, Inference, and Learning Algorithms", Cambridge Univ. press 2003, http://www.inference.org.uk/mackay/itila/book.html
Description of exams and course work
The achievement of the competencies given in section learning outcome is tested exemplarily at least to the given cognition level using presentations independently prepared by the students. The exam of 25 minutes consists of the presentation and a subsequent discussion.
Aspects of evaluation are especially:
- Illustrative and clear presentation of a topical research field within a scientific talk
- Answering questions on the scientific content of the talk
Participation in the exercise classes is strongly recommended since the exercises prepare for the problems of the exam and rehearse the specific competencies.
The exam may be repeated at the end of the semester.
Current exam dates
Currently TUMonline lists the following exam dates. In addition to the general information above please refer to the current information given during the course.
|Exam to Introduction to Neuronal Networks for Physicists
|Dummy-Termin, Leistung wird im Rahmen der Veranstaltung erbracht. // Dummy date. Course work during the semester.