Third-cycle subject: Computer Science
1) Topology and Geometry in Machine Learning
Developing the theoretical underpinnings of machine learning algorithms is crucial for making learning robust, data-efficient, and explainable. Understanding and exploiting geometric and topological properties of data is one of the challenges associated to this topic.
One position is focused on applying geometric and topological concepts and techniques in order to (i) analyze modern generative models, such as Variational Autoencoders and Generative Adversarial Networks, and (ii) improve their performance by enforcing geometric and topological priors.
The successful applicants will be expected to have a background and interest in Differential and Computational Geometry and Applied Topology, as well as a strong interest in Machine Learning and good programming skills.
2) Reinforcement learning, perception, and control
The second position will focus on integrating machine learning, perception, and control. We look into reinforcement learning and model predictive control for addressing problems such as robotic interaction with soft objects. The successful applicants will be expected to have a background and interest in automatic control and robotics, as well as a strong interest in Machine Learning and good programming skills.
Supervision: The doctoral students will be supervised by Danica Kragic Jensfelt
Further details:
Doctoral students in Topology, Geometry and Reinforcement learning at KTH Royal Institute of Technology