RL_project: Comparison of Q-Learning and Double Q-Learning

Het grote versterkende leren project

For the figures see the results.ipynb notebook.

Requirements

Python >= 3.8.5 with:

• numpy >= 1.19.2
• matplotlib >= 3.3.4
• gym >= 0.19.0
• tqdm >= 4.36.0

Overview of files

Environments (codebase/environments/)

Contains the implementations of the Gym environments, with the following files:

• codebase/environments/stochasticEnv.py: The base stochastic Gridworld Gym environment implementation
• codebase/environments/grid_world.py: The deterministic, stochastic and left-right Gridworld implementations used for the paper
• codebase/environments/left_right.py: The Left-Right problem implementatation

Models (codebase/models/)

Contains the implementation of Q-Learning and Double Q-Learning (codebase/models/q_learning.py) and the behaviour policies (codebase/models/policies.py)

Experiments (codebase/experiments.py)

Finally, codebase/experiments.py contains the code for the different experiments we ran to obtain the results in the paper