Agenda

This tutorial hosted at EASSS 2024, will be composed of a theoretical lesson on Reinforcement Learning (RL), followed by a laboratory mainly focusing on RL environments setup.

Theoretical Session (10:00 - 11:30)

Description

The tutorial will begin by covering fundamental RL concepts such as agents, environments, and learning paradigms. We will then discuss why deep learning is increasingly employed to solve complex, high-dimensional decision-making challenges.

📑 Theoretical Session Slides

Practical Session (12:00 - 13:30)

Description

The hands-on component will guide participants through setting up a training environment tailored for RL agents. This part of the tutorial will include a practical introduction with a grid world example to the OpenAI’s Gymnasium library and references the StableBaselines3 library that implement deep RL algorithms.

🛠️ Practical Session Slides

📦 Tutorial Code

Environment Setup

To create a conda environment and install the necessary libraries, use the following commands:

conda env create --name rleasss
conda activate rleasss
pip install stable-baselines3 tensorboard

Or use the requirements.txt file:

conda env create --name rleasss
conda activate rleasss
pip install -r requirements.txt

Or use the attached environment.yml file:

conda env create -f environment.yml
conda activate rleasss

If you use venv, do the same steps and use the requirements.txt file or the manual installation:

python -m venv myenv
source myenv/bin/activate
pip install -r requirements.txt

However, with MacOS some issues may arise based on the numpy version. In that case downgrade numpy to a version lower than 2.0:

pip install "numpy<2"

I look forward to a productive and insightful tutorial. Please ensure you have the necessary environment set up before the practical session begins.