Logic offers a powerful framework to understand complex systems. It is a cornerstone of formal methods — fundamental tools to reason about strong guarantees of complex systems. Learning is an ubiquitous paradigm, whose applications seem limitless. It is mostly based on statistical approaches with unparalleled efficiency. This project participates in a blooming effort to combine the formal reasoning from logic and the power of learning. On the one hand, introducing learning in formal methods techniques has the potential to solve many performance issues while preserving hard guarantees. On the other hand, logic provides mathematical foundations to assess the quality of learning processes and outputs, and to guarantee their safety and robustness, in an era where learning and AI are at the core of critical applications.