Artificial Intelligence meets Formal Methods

Organization

The seminar is jointly organized with Prof. Neider (TU Dortmund University) and Prof. Zeume (Ruhr University Bochum). There will be an initial meeting on October 11, 10 am (c.t.), room see Moodle. In the initial meeting, we will present and assign the topics. In the first week after the semester, there will be a block seminar in which students present their results.

More up-to-date information will be made available in Moodle; please register there.

Content

In this seminar we will explore recent and classic advances in the interface between the fields Artificial Intelligence and Formal Methods. For example, we investigate

(1) how formal descriptions and models can be learned from data,

(2) the expressive power of machine learning models,

(3) how to learn automata using machine learning methods,

(4) how to verify neural networks using formal methods,

(5) how to use automata in reinforcement learning, or

(6) how to solve SAT problems using machine learning.

The goal of the seminar is to understand how to combine methods from both areas in order to get, for instance, explainability in machine learning, to understand (from a theoretical point of view) the borders of machine learning, or to improve machine learning by using formal methods.

Most topics will be of rather theoretical nature, but there will also be some more practical topics.