Reinforcement Learning Coursework

This repository contains coursework completed for the PhD-level course
Reinforcement Learning at Linköping University.

The course evaluation is based on four programming projects, written reports, assignments, and reflective journals. There is no written exam.

Course information

• Course page: https://farad59.gitlab-pages.liu.se/reinforcement_learning_phd_course/en/
• Level: PhD
• Evaluation: Projects, assignments, reflective journals

Repository structure

projects/

Contains the four mandatory course projects:

• Project 1 – Basic reinforcement learning algorithms
  (SARSA, Expected SARSA, Q-learning, Double Q-learning, Monte Carlo)
  Applied to Blackjack.

• Project 2 – Normalized Advantage Functions (NAF)
  Applied to the inverted pendulum.

• Project 3 – Deep Deterministic Policy Gradient (DDPG) and
  Soft Actor-Critic (SAC)
  Applied to the inverted pendulum.

• Project 4 – Portfolio project (PPO-based methods)
  Includes experiments with PPO and masked PPO on custom environments, including Connect Four and an exploratory adaptation to Hanabi. A written report summarizes results, design choices, and observations.

instructions/

Official project descriptions provided by the course instructors.

project4_report/

Final report and figures for Project 4.

Notes on implementation

• Projects 1–3 were completed by filling in provided code skeletons, as required by the course.
• Project 4 was an open-ended portfolio project designed and implemented by the student.
• Some experiments build on existing libraries (e.g., Stable-Baselines3), with adaptations, analysis, and discussion documented in the report.

Academic integrity

This repository contains original coursework produced by the repository author

The repository is shared for documentation and portfolio purposes only.