{"external_id":{"arxiv":["2405.12654"]},"author":[{"full_name":"Köhler, Dominik","last_name":"Köhler","first_name":"Dominik"},{"last_name":"Heindorf","full_name":"Heindorf, Stefan","id":"11871","orcid":"0000-0002-4525-6865","first_name":"Stefan"}],"language":[{"iso":"eng"}],"oa":"1","department":[{"_id":"760"}],"type":"preprint","user_id":"11871","year":"2024","title":"Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/2405.12654"}],"date_updated":"2024-05-26T19:04:55Z","status":"public","date_created":"2024-05-26T18:49:59Z","abstract":[{"lang":"eng","text":"Graph Neural Networks (GNNs) are effective for node classification in\r\ngraph-structured data, but they lack explainability, especially at the global\r\nlevel. Current research mainly utilizes subgraphs of the input as local\r\nexplanations or generates new graphs as global explanations. However, these\r\ngraph-based methods are limited in their ability to explain classes with\r\nmultiple sufficient explanations. To provide more expressive explanations, we\r\npropose utilizing class expressions (CEs) from the field of description logic\r\n(DL). Our approach explains heterogeneous graphs with different types of nodes\r\nusing CEs in the EL description logic. To identify the best explanation among\r\nmultiple candidate explanations, we employ and compare two different scoring\r\nfunctions: (1) For a given CE, we construct multiple graphs, have the GNN make\r\na prediction for each graph, and aggregate the predicted scores. (2) We score\r\nthe CE in terms of fidelity, i.e., we compare the predictions of the GNN to the\r\npredictions by the CE on a separate validation set. Instead of subgraph-based\r\nexplanations, we offer CE-based explanations."}],"citation":{"chicago":"Köhler, Dominik, and Stefan Heindorf. “Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks.” <i>ArXiv:2405.12654</i>, 2024.","ieee":"D. Köhler and S. Heindorf, “Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks,” <i>arXiv:2405.12654</i>. 2024.","apa":"Köhler, D., &#38; Heindorf, S. (2024). Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks. In <i>arXiv:2405.12654</i>.","ama":"Köhler D, Heindorf S. Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks. <i>arXiv:240512654</i>. Published online 2024.","short":"D. Köhler, S. Heindorf, ArXiv:2405.12654 (2024).","bibtex":"@article{Köhler_Heindorf_2024, title={Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks}, journal={arXiv:2405.12654}, author={Köhler, Dominik and Heindorf, Stefan}, year={2024} }","mla":"Köhler, Dominik, and Stefan Heindorf. “Utilizing Description Logics for Global Explanations of Heterogeneous  Graph Neural Networks.” <i>ArXiv:2405.12654</i>, 2024."},"_id":"54448","publication":"arXiv:2405.12654"}