This code demonstrates our implementation of SARSA for a deterministic FrozenLake task and serves as additional information to go alongside the "Building a complete Rl system" lecture. The lecture is delivered as part of the Reinforcement Learning (2020) course at the University of Edinburgh.
To run the python code, you will need OpenAI Gym which can be installed using
pip install gym
Further dependencies are Matplotlib and NumPy which can be installed with
pip install matplotlib
pip install numpy
We strongly suggest you to go through the jupyter notebook file of this demonstration in your own time! This includes further information and explanations to understand this demonstration and learn more about good practises in RL evaluation. You can directly view the notebook here or run it yourself with jupyter. To install the required software and run the notebook provided with the code, follow the instructions on the Jupyter mainpage.
All code is written in Python3 and provided as separate scripts and all-together in one Jupyter notebook available here with further explanations.
This lecture was delivered by the TA team consisting of
- Arrasy Rahman arrasy.rahman@ed.ac.uk
- Filippos Christianos f.christianos@ed.ac.uk
- Lukas Schäfer l.schaefer@ed.ac.uk
Course organisers are
- Stefano Albrecht s.albrecht@ed.ac.uk
- Pavlos Andreadis pavlos.andreadis@ed.ac.uk