| Objectives |
This subject focuses on modern AI techniques like Deep Learning which is a major driving force in AI and provides many of its success stories. The fundamental deep learning problem is introduced and its solutions are discussed, including proper activation functions, specific network architectures such as residual neural networks as well as highway networks. Additionally, learning and regularization techniques as well as unsupervised methods such as variational autoencoders and generative adversarial networks (GANs) are presented. Students will also learn about recurrent neural nets like long short-term memory (LSTM) with applications in text and speech processing and reinforcement learning.
Another subject matter is perception with a focus on vision. Moreover, the theoretical aspects of machine learning like statistical learning theory, estimation theory, or Bayes techniques are presented. These topics are completed by probabilistic models covering, among others, probabilistic graphical models, Bayesian networks, Kalman filters, and hidden Markov models.
|
