This Seminar/Project can be used for the courses (see TUGonline):

• 708.424 Seminar/Projekt Computational Intelligence
• 708.041 Seminar/Projekt Machine Learning and Neuroinformatics
• 708.415 Seminar/Project Machine Learning & Neuroinformatics/Brain-Computer Interfacing

Content of the seminar/project: Brain-Inspired Artificial Intelligence

Within the last decade, deep learning has revolutionized Artificial Intelligence research. Deep learning is based on neural networks, which are themselves to some extent a brain-inspired model. However, in order to increase their capabilities, neural networks have often been extended using ideas that were inspired from our knowledge of the brain.

In this Seminar/Project, students will work in small groups of 2-3 persons on software projects where they implement brain-inspired AI models based on scientific literature. We will discuss the progress throughout the semester and students present their work at the end.

Students are expected to have visited the courses Neural Networks or Principles of Brain Computations (or similar courses) prior to this Seminar/Project. In parallel, there will be a Seminar (Seminar Computational Intelligence A) which will tackle the same topic and is a good complement to this Seminar/Project.

Topics for Projects include (but are not limited to):

• Rapid learning and task solving with neural networks
• Deep Reinforcement Learning
• Neural Networks with brain-inspired memory-components
• Spiking encoder/decoder architectures
• Variational autoencoders for unsupervised learning
• Learning from imitation
• Spiking convolutional networks
• Energy-based models