Functional Synthesis: Formal Methods meets Machine Learning
When: Monday, October 9, 2023, 15:30 (BST).
Where: In person and Zoom online.
Topic: Functional Synthesis: Formal Methods meets Machine Learning.
Speaker: Maximilian Prokop, PhD candidate.
Abstract
Partial exploration and on-the-fly algorithms are crucial to tackle otherwise completely intractable problems. The performance of such algorithms depends heavily on the exploration heuristic. In this talk, we demonstrate how one can harness the power of machine learning for this purpose and thus explore semantically labelled state spaces more efficiently. We will explain all ideas with the example of parity games in LTL synthesis, as we also did in our paper “Guessing Winning Strategies in LTL Synthesis using Semantic Learning” from this years iteration of CAV.
Short Bio
Maximilian Prokop is a first year joint PhD candidate of Javier Esparza at the TU Munich and Jan Kretinsky at the Masaryk University of Brno. His major research interest is the applicability of machine learning in the realm of formal methods which up to now manifested developing ML-based exploration heuristics for LTL Synthesis.