Abstract: Graph Neural Networks (GNNs) are often used to learn transformations of graph data. While effective in practice, such approaches make predictions via numeric manipulations so their output cannot be easily explained symbolically. We propose a new family of GNN-based transformations of graph data that can be trained effectively, but where all predictions can be explained symbolically as logical inferences in Datalog—a well-known rule-based formalism. In particular, we show how to encode an input knowledge graph into a graph with numeric feature vectors, process this graph using a GNN, and decode the result into an output knowledge graph. We use a new class of monotonic GNNs (MGNNs) to ensure that this process is equivalent to a round of application of a set of Datalog rules. We also show that, given an arbitrary MGNN, we can automatically extract rules that completely characterise the transformation. We evaluate our approach by applying it to classification tasks in knowledge graph completion.
One-sentence Summary: We propose a new family of graph neural network-based transformations of graph data that can be trained effectively and where all predictions can be explained symbolically as logical inferences in Datalog.
Supplementary Material: zip