Seminar: Reinforcement Learning - Markov Decision Processes (MDPs) Report & Presentation

A repository contains resources related to a seminar offered at Paderborn University (Summer-2023) on Reinforcement Learning, focusing on Markov Decision Processes (MDPs). It includes a comprehensive report and a presentation (PPT) summarizing key concepts, Finite MDPs vs Infinite MDPs, and different methods for solving Finite and Infinite MDPs Problem.

Contents:

• Report detailed document.
• Presentation slides summarizing the key points covered in the seminar report.

Abstract:

Reinforcement Learning (RL) is a subfield of machine learning (ML) that focusses on training agents to make sequential decisions in dynamic environments to maximise cumulative rewards. Markov Decision Processes (MDPs) provide a fundamental framework for modelling and solving RL problems. This report aims to explore the concepts and implications of MDPs in the context of RL, specifically focusing on finite and infinite MDPs.

Keywords: Reinforcement Learning · Markov Decision Processes · Finite MDPs · Infinite MDPs.