Course Content & Topics
Theoretical and practical aspects of computing and learning with neural networks. Investigation of the most commonly used algorithms for deep learning. From the practical perspective, various learning algorithms and types of neural networks will be implemented and applied to artificial and real-world problems. A list of the topics that will be covered is as follows:
- Theoretical background on machine learning
- Feedforward neural networks
- Training methods, optimization algorithms
- Regularization, generalization
- Convolutional neural networks
- Recurrent neural networks (LSTMs, GRUs, etc.)
- Deep generative models
- Variational autoencoders
- Generative adversarial networks
- Denoising diffusion probabilistic models
- Attention & Transformers
Learning Objectives
After positive completion of the course, students will be able to:
- Understand and apply the fundamental concepts of learning principles to implement commonly used architectures in deep learning and to develop novel architectures.
- Design and train complex deep neural networks in supervised and unsupervised learning scenarios which requires a thorough theoretical and practical understanding of the algorithms.
- Identify relevant and important features, benefits and limitations of different neural network models and algorithms, e.g., with respect to their practical, generalization- and regularization abilities.
- Apply state-of-the-art deep learning methods to different problems and to analyze, monitor and visualize the models’ performance and limitations.
Location & Time
Location: HS Thermoprozesstechnik (05ME01124) during October. Starting in November, lectures are planned to be held in HS 3 Studienzentrum (35SZ02211).
Time: Tuesdays & Thursdays 10:00 – 12:00. Detailed schedule can be found here.
Grading
* Continuous assessment and written exam: Details will be presented in the first lecture unit.
* Task assignments: Several practical assignments have to be implemented. For each assignment a written report and/or slides have to be submitted. Details will be presented in the first lecture unit.
* Grading scheme: 0-49.9 Points (5), 50-62.4 Points (4), 62.5-74.9 Points (3), 75-87.4 Points (2), 87.5-100 Points (1)
(With an overall score of up to 75%, an additional oral performance review MAY (!) also be required if the positive performance record is not clear. In this case, you will be informed as soon as the grades are released. If you have already received a grade via MU online, you will not be invited to another oral performance review.)
Prerequisites
- Formal prerequisite: Introduction to Machine Learning VU (“190.018”) or L+P (“190.012” and “190.013”).
- Recommended prerequisites: Good Python programming skills, Fundamentals of Probability Theory, Basic Algebra & Vector Calculus.
Literature
– Ian Goodfellow, Yoshua Bengio and Aaron Courville, “Deep Learning”, 2016.
– Christopher M Bishop, “Pattern Recognition and Machine Learning”, 2006.