Bridging max graph neural networks and datalog with negation
We consider a general class of data transformations based on Graph Neural Networks (GNNs), which can be used for a wide variety of tasks. An important question in this setting is characterising the expressive power of these transformations in terms of a suitable logic-based language. From a practica...
| Autori principali: | , |
|---|---|
| Natura: | Conference item |
| Lingua: | English |
| Pubblicazione: |
IJCAI Organization
2024
|
| _version_ | 1826316103958134784 |
|---|---|
| author | Tena Cucala, D Cuenca Grau, B |
| author_facet | Tena Cucala, D Cuenca Grau, B |
| author_sort | Tena Cucala, D |
| collection | OXFORD |
| description | We consider a general class of data transformations based on
Graph Neural Networks (GNNs), which can be used for a
wide variety of tasks. An important question in this setting is
characterising the expressive power of these transformations
in terms of a suitable logic-based language. From a practical
perspective, the correspondence of a GNN with a logical theory can be exploited for explaining the model’s predictions
symbolically. In this paper, we introduce a broad family of
GNN-based transformations which can be characterised using Datalog programs with negation-as-failure, which can be
computed from the GNNs after training. This generalises existing approaches based on positive programs by enabling the
learning of nonmonotonic transformations. We show empirically that these GNNs offer good performance for knowledge
graph completion tasks, and that we can efficiently extract
programs for explaining individual predictions. |
| first_indexed | 2024-09-25T04:29:58Z |
| format | Conference item |
| id | oxford-uuid:c54d402c-bda1-4ee9-8a9d-f665b662a3e1 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:37:43Z |
| publishDate | 2024 |
| publisher | IJCAI Organization |
| record_format | dspace |
| spelling | oxford-uuid:c54d402c-bda1-4ee9-8a9d-f665b662a3e12024-12-03T11:55:56ZBridging max graph neural networks and datalog with negationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:c54d402c-bda1-4ee9-8a9d-f665b662a3e1EnglishSymplectic ElementsIJCAI Organization2024Tena Cucala, DCuenca Grau, BWe consider a general class of data transformations based on Graph Neural Networks (GNNs), which can be used for a wide variety of tasks. An important question in this setting is characterising the expressive power of these transformations in terms of a suitable logic-based language. From a practical perspective, the correspondence of a GNN with a logical theory can be exploited for explaining the model’s predictions symbolically. In this paper, we introduce a broad family of GNN-based transformations which can be characterised using Datalog programs with negation-as-failure, which can be computed from the GNNs after training. This generalises existing approaches based on positive programs by enabling the learning of nonmonotonic transformations. We show empirically that these GNNs offer good performance for knowledge graph completion tasks, and that we can efficiently extract programs for explaining individual predictions. |
| spellingShingle | Tena Cucala, D Cuenca Grau, B Bridging max graph neural networks and datalog with negation |
| title | Bridging max graph neural networks and datalog with negation |
| title_full | Bridging max graph neural networks and datalog with negation |
| title_fullStr | Bridging max graph neural networks and datalog with negation |
| title_full_unstemmed | Bridging max graph neural networks and datalog with negation |
| title_short | Bridging max graph neural networks and datalog with negation |
| title_sort | bridging max graph neural networks and datalog with negation |
| work_keys_str_mv | AT tenacucalad bridgingmaxgraphneuralnetworksanddatalogwithnegation AT cuencagraub bridgingmaxgraphneuralnetworksanddatalogwithnegation |