Martin Kurečka

Safe and Reliable Reinforcement Learning Through Constraints and Learnt Models

Lector: Petr Novotný

Institution: Masaryk University, Faculty of Informatics

Field: Fundamentals of Computer Science

 

About the project

This project focuses on Reinforcement Learning (RL), a method for training autonomous agents such as cars and robots. Although RL has found successful applications in areas such as robotics and finance, the reliability of trained agents remains a key concern. The project has three main goals: to improve ways problems can be modeled as RL tasks, to formally verify the long-term "intentions" of agents, and to develop new monitoring techniques to detect deviations from these intentions on the fly.

 

Why science?

I've always loved solving math problems and learning new things, and science offers countless opportunities to do both. Unlike the commercial sphere, where the choice of a problem is often subject to the potential for profit, science provides the freedom to explore problems just out of curiosity.

 

What do I like most about Brno

Personally, I think Brno has a lot of great communities, whether it's sports, music gatherings, board games, science or culture. It's a big enough city to have interesting things going on, but small enough that you don't get lost and people know each other.