TL;DRAbstract
The chess endgames are a perfect test domain for Machine Learning algorithms. Reciprocally, Machine Learning o ers e cient methods to analyse and compress chess tablebases. Such databases consist of informations at a low level of description and are often huge, what makes them unpractical. Our goal is to reduce the size of the King and Rook against King (KRK) database without information loss with the help of knowledge-based compression. Some studies already tried in the case of the KRK endgame to use speci c patterns in order to build a knowledge with a higher level of description. Grouping such patterns together, we transform the primary tablebase in a more compact form containing only values of these attributes. The representation of this new database as a decision tree simplies its description, enabling a better compression rate than standard mechanical methods.
Chat with Paper
AI Agents for this Paper
The chess endgames are a perfect test domain for Machine Learning algorithms. Reciprocally, Machine Learning o ers e cient methods to analyse and compress chess tablebases. Such databases consist of informations at a low level of description and are often huge, what makes them unpractical. Our goal is to reduce the size of the King and Rook against King (KRK) database without information loss with the help of knowledge-based compression. Some studies already tried in the case of the KRK endgame to use speci c patterns in order to build a knowledge with a higher level of description. Grouping such patterns together, we transform the primary tablebase in a more compact form containing only values of these attributes. The representation of this new database as a decision tree simplies its description, enabling a better compression rate than standard mechanical methods.
Keywords
Chat
Click to start Chat