Python XCS Framework

Assigned Computer Science Games Engineering Bachelor Master Project Thesis Games Research Lab

Background

Artificial Neural Networks are commonly used for machine learning. Many frame works exist, that enable machine learning newcomers and veterans alike to solve many learning tasks without needing to program anything themselves. A downside of artificial neural networks lies in them being a black box as it is hard to understand what actually happens inside them as they learn. Learning classifier systems (LCS) offer a human-understandable rule-based approach to machine learning but there are fewer frameworks and no de-facto standard implementation. Much research for machine learning is done in Python and there exist multiple excellent frameworks. To enable an easier entry into the practical use of learning classifier systems, the goal of this project is to provide a framework that implements the basic methods some of their variants of learning classifier systems.

Tasks

• Research the current trends in learning classifier systems
• Implement an eXtended Classifier System (XCS) with original bitstring representation
• Add support for real-value inputs
• Identify and implement important additional functions (rule-combination, function-approximation…)
• Compare the implemented methods in a well-known reinforcement learning task, such as the CartPole or MountainCar environment

Contact Persons at the University Würzburg