This repository gives a brief introduction to understand Markov Decision Process (MDP). The following material is part of Artificial Intellegence (AI) class by Phd. Carlos A. Lara Álvarez in Center for Research In Mathematics-CIMAT (Spring 2019).
Two exercises are given.
- Wikipedia exercise - Example of a simple MDP with three states (green circles) and two actions (orange circles), with two rewards (orange arrows).
- World-grid - Example of a MDP with 13 stages (white boxes) and four actions (up, right, down, left), with two rewards (green box and red box)
The algorithm consist on a Policy Iteration. For an explanation of policy Iteration I highly recommend to read "Reinforcement Learning: An Introduction" by Richard Sutton.
Policy Iteration uses a policy evaluation (evaluate a given policy) and policy improvement (finds the best policy).
The scripts were coded on MATLAB.
The code in this repository, including all code samples in the notebooks listed above, is released under the MIT license. Read more at the Open Source Initiative.