Lecturing at KTH Royal Institute of Technology
- Artificial Intelligence, Master-level course (2012, 2013, 2014, 2015)
- Reinforcement Learning, PhD.-level course (2017)
The course covers problem solving with search algorithms, heuristics and games, knowledge representation (logic), representing uncertain knowledge and reasoning (Bayesian networks), decision and utility theory, ommunication between agents, and models for probabilistic language processing. Further are covered examples of using artificial intelligence methods in computer vision, robotics, etc will be given.
This course is about Reinforcement Learning (RL), the third type of learning besides Supervised Learning and Unsupervised Learning. In RL, the paradigm is to try out different options and learn from feedback which options are best to reach a goal. RL has its roots both in biologically and psychologically inspired learning approaches and in control. The course provides a short and basic introduction to classical RL and advanced RL topics. Each lecture comes with a practical task.
For this course, I received funding from the PhD. student council of the School of Computer Science and Communication at KTH.
Supervision at KTH Royal Institute of Technology
Filip Domazet, Bachelor Project 2012: Learning structure and control of robotic hands, genetic algorithms, Artificial neural networks, physical simulation
Akshaya Thippur Sridatta, PhD. student with Prof. Dr. P. Jensfelt and Prof. Dr. C.H. Ek 2016–2017: Probabilistic reasoning, structured prediction and classification problems, patial relation calculi, spatial perception and reasoning, scene understanding
Alejandro Marzinotto, PhD. student with Prof. Dr. D. Kragic 2016–2017: Robot control with topological spaces, manipulation of deformable objects, tracking and representa- tion of deformable objects
Mia Kokic, PhD. student with Prof. Dr. D. Kragic 2016–: Object affordances for task-specific grasping, 3D perception, semantic segmentation, deep learning
Isac Arnekvist, Master Project 2017: Deep Reinforcement learning, parallel and distributed Reinforcement learning, robotic and non- prehensile manipulation
Isac Arnekvist, PhD. student with Prof. Dr. D. Kragic 2017–: Deep Reinforcement learning for learning dynamic systems, deep learning of probabilistic models, robotic and non-prehensile manipulation