Advanced Deep Learning for Robotics

What it is about

In this module you deepen advanced deep learning methods with a focus on robotic applications and deep reinforcement learning. You learn both theoretical foundations of modern network architectures and probabilistic methods as well as their practical implementation in simulations and projects for robotics tasks.

What you will be able to do

• Understand advanced architectures of neural networks
• Apply Bayesian Deep Learning and uncertainty estimation
• Know and use generative models (VAE, GAN)
• Data-efficient learning: apply transfer- and semi-supervised methods
• Understand state-of-the-art Deep Reinforcement Learning algorithms
• Practically implement and train DRL methods in robotic simulations

What the module consists of

• LectureConveying the theoretical foundations of Advanced Deep Learning and Deep Reinforcement Learning
• Practical sessionsPractical implementation, training and testing of DRL methods in simulations
• Semester project (teams of 2)Application and deepening through a semester-long project with weekly presentations and tutoring

Teaching method

• LectureDetailed theoretical presentation and discussion of current research
• Reading AssignmentsDeepening through selected chapters and current conference papers
• Practical exercises and trainingGaining practical experience with implementation and training of DRL methods
• Project supervision with weekly presentationsIterative work on a real problem and continuous feedback