Constraint Acquisition - A Tutorial on Learning Constraint Models

Abstract

Dimos Tsouros presents a tutorial on Learning Constraint Models. Constraint Programming (CP) is a powerful paradigm for solving complex combinatorial problems, but its adoption is often hindered by the expertise required for modeling. Constraint Acquisition (CA) aims to mitigate this bottleneck by semi-automating the modeling process. This tutorial will provide a comprehensive introduction to CA, covering both passive and interactive learning approaches. For passive acquisition, we will explore CA techniques for learning constraint models from datasets of existing solutions and non-solutions. We will discuss different approaches that focus on learning fixed-arity or global constraints, handling noise, and generalizing the learned models to handle varying problem instances. We will also review interactive CA techniques, highlighting the recent integration of statistical Machine Learning methods that enhance efficiency by reducing the number of queries needed. During the tutorial, state-of-the-art CA tools implemented in the open-source CPMpy modeling language will be demonstrated. Finally, we will discuss current challenges and future directions in constraint acquisition research