This repository contains the source code for the numerical experiments presented in the paper "Analyzing Policy Entropy of Reinforcement Learning Agents for Personalization Tasks".
Install the requirements via pip install -r requirements.txt
.
Run the experiments via python -m run_experiment -c config
, where config
is a configuration file in ./configs/
directory.
The available values are {config_mnist, config_cifar10, config_spotify, config_recogym, config_personalization}
, which could be specified to recreate each of the presented numerical experiments.
Optionally, a custom experiment can be set up by changing or adding new configuration file.
All previously performed experiments are stored in ./data/
directory and can be recreated by loading via python -m run_experiment -l exp_name
, where exp_name
is the name of the experiment as it is saved in ./data/
.
run_experiment.py
--- set up and run the experimentagent.py
--- set up selected RL agentsenvironment.py
--- create the specified environmentenvironments/
--- data required to set up various environmentsconfigs/
--- configuration files for the experimentsdata/
--- store data from previously run experimentsimages/
--- plots of various results from experimentsvisualization.py
--- save/load the experiment data, plot the results
This project is licensed under the MIT License.