A neuro-symbolic approach to computing alignments in Causal Nets

In the process mining, one of the computationally expensive tasks is computing alignments, i.e., mapping a trace representing the execution of a business process to a model of the business process. This is especially pronounced when using Causal Nets, an expressive model representation requiring making multiple decisions that are not directly reflected in a trace. The goal of the thesis would be to extend an exact approach (e.g., the A* algorithm) with a learned heuristic to efficiently prioritize the algorithm so it performs well in real-world scenarios.