Hello, I am Tim and I am researching unsupervised reinforcement learning. I studied computer science at KIT and I am passionate about machine learning. If you are interested in how autonomous agents can learn about the world without any supervision, apply for one of the open positions below or just send me an email with a topic that interest you.
Looking for a research assistant (HiWi) job? Apply to one of the open positions or message me via Email.
Looking for a bachelor / masters thesis? Apply to one of the open positions or message me via Email with a broad topic idea.
Reinforcement Learning: Efficient Exploration (research assistant / thesis)
With reinforcement learning one can train an agent (e.g. a robot) to execute a number of actions (e.g. moving to a goal) to solve a task. However, reinforcement learning is data-intensive and an agent usually needs a lot of time to adapt to a new task. If the agent could learn beforehand how to world is structured, it seems natural that the agent should be able to adapt faster.
Thus, efficiently exploring the world to learn about it, is necessary.
What you will do:
- Research new approaches on efficient exploration based on the state-of-the-art
- Implement your algorithms into reinforcement learning agents
- Improve over existing methods and contribute the field of unsupervised reinforcement learning
What you can expect:
- Bi-weekly meetings with supervisor
- Discussion on the current state-of-the-art with supervisor and other students
- Coding support (e.g. given baseline code)
- No non-research related tasks
- Possibility to continue research with a bachelors/masters thesis
- Flexible work hours (i.e. study time/exams get priority)
What you should bring:
- Being able to write code in Python
- Already implemented something in either PyTorch or Tensorflow
- Visited a course on neural networks, probability theory or reinforcement learning
- Speak either German or English
To apply: Send me an email (joseph∂fzi de) with your résumé and grade table (Notenauszug) and ideally, if you can send me a link to the coolest paper you ever read. No need for a cover letter.
Bachelor / Master thesis
Send me an email with what interest you or an idea. Some of my ideas:
- What happens beyond the exploration boundary? When an agent explores it will advance into unknown regions of the environment. However, how the agent interacts with this not yet explored part of the environment is unclear. Will it still show sensible behavior or will it deterioate?
If you are interested in one of these topics, message mia via email (joseph∂fzi de) and we can discuss the topic further to see, whether this is something that is suitable for you as a thesis topic.