Advanced Machine Learning for Physics and Scientific Discovery

• Lecturer: Florian Marquardt
• 5 ECTS credit points
• One 90min lecture per week
• Monday 18:00 - 19:30, lecture hall F
• written exam

We will describe advanced modern methods of artificial intelligence and their potential application to artificial scientific discovery, in physics and other fields. This includes:

• representation learning (including deep variational autoencoders etc.)

• active learning (how a neural network can choose suitable training samples on its own)

• reinforcement learning and optimization methods

• graph neural networks

• generative neural networks (learning to sample from an observed statistical distribution)

• transformers and other attention-based methods

• advanced concepts from information science and statistics (e.g. mutual information)

• automated program discovery

• applications in quantum physics, statistical physics, dynamical systems

For information (in German) on the bachelor theory colloquium, please refer to  ► exam info

Seminars and lectures of past semesters can be found in our ► teaching archive