In the first lecture, we have seen that computing the output of extension semantics (preferred, stable) is very hard. In this lecture, we explain how machine learning can help in building heuristics.

5A.1 Background on Graph Neural Networks

5A.2 Classification of arguments for credulous acceptability (Kuhlmann and Thimm)

5A.3 Identified Problems

5A.4 Improving GCN for credulous/skeptical acceptability (Malmqvist et al.)

5A.5 Using GNNs to select solvers (Klein et al. 2022)

5A.6 Conclusion