YAD: A Learning-based Inductive Logic Programming Tool

Inductive Logic Programming (ILP) learns first-order (Horn-clause) rules from relational data expressed as background knowledge plus positive and negative examples, but practical ILP is often dominated by expensive search over large hypothesis spaces. YAD is an open-source, C#-based ILP tool that implements a new bottom-up induction algorithm built on inverse resolution, designed to reduce learning time by prioritizing rule candidates that are more likely to yield relevant generalizations rather than relying on purely random exploration. The tool provides a graphical interface for data loading and parameter configuration, supports rule induction, pruning to improve generality and reduce overfitting, and evaluation on held-out data using standard metrics (e.g., precision, recall, accuracy, and F-measure). YAD is intended for discovering relationships in multi-relational domains such as semantic web ontology alignment, bioinformatics, and social network analysis.