{"_id":"29290","title":"EvoLearner: Learning Description Logics with Evolutionary Algorithms","department":[{"_id":"574"}],"user_id":"11871","citation":{"short":"S. Heindorf, L. Blübaum, N. Düsterhus, T. Werner, V.N. Golani, C. Demir, A.-C. Ngonga Ngomo, in: WWW, ACM, 2022, pp. 818–828.","chicago":"Heindorf, Stefan, Lukas Blübaum, Nick Düsterhus, Till Werner, Varun Nandkumar Golani, Caglar Demir, and Axel-Cyrille Ngonga Ngomo. “EvoLearner: Learning Description Logics with Evolutionary Algorithms.” In <i>WWW</i>, 818–28. ACM, 2022.","bibtex":"@inproceedings{Heindorf_Blübaum_Düsterhus_Werner_Golani_Demir_Ngonga Ngomo_2022, title={EvoLearner: Learning Description Logics with Evolutionary Algorithms}, booktitle={WWW}, publisher={ACM}, author={Heindorf, Stefan and Blübaum, Lukas and Düsterhus, Nick and Werner, Till and Golani, Varun Nandkumar and Demir, Caglar and Ngonga Ngomo, Axel-Cyrille}, year={2022}, pages={818–828} }","apa":"Heindorf, S., Blübaum, L., Düsterhus, N., Werner, T., Golani, V. N., Demir, C., &#38; Ngonga Ngomo, A.-C. (2022). EvoLearner: Learning Description Logics with Evolutionary Algorithms. <i>WWW</i>, 818–828.","ama":"Heindorf S, Blübaum L, Düsterhus N, et al. EvoLearner: Learning Description Logics with Evolutionary Algorithms. In: <i>WWW</i>. ACM; 2022:818-828.","mla":"Heindorf, Stefan, et al. “EvoLearner: Learning Description Logics with Evolutionary Algorithms.” <i>WWW</i>, ACM, 2022, pp. 818–28.","ieee":"S. Heindorf <i>et al.</i>, “EvoLearner: Learning Description Logics with Evolutionary Algorithms,” in <i>WWW</i>, 2022, pp. 818–828."},"date_created":"2022-01-12T10:22:53Z","status":"public","year":"2022","oa":"1","abstract":[{"text":"Classifying nodes in knowledge graphs is an important task, e.g., predicting\r\nmissing types of entities, predicting which molecules cause cancer, or\r\npredicting which drugs are promising treatment candidates. While black-box\r\nmodels often achieve high predictive performance, they are only post-hoc and\r\nlocally explainable and do not allow the learned model to be easily enriched\r\nwith domain knowledge. Towards this end, learning description logic concepts\r\nfrom positive and negative examples has been proposed. However, learning such\r\nconcepts often takes a long time and state-of-the-art approaches provide\r\nlimited support for literal data values, although they are crucial for many\r\napplications. In this paper, we propose EvoLearner - an evolutionary approach\r\nto learn ALCQ(D), which is the attributive language with complement (ALC)\r\npaired with qualified cardinality restrictions (Q) and data properties (D). We\r\ncontribute a novel initialization method for the initial population: starting\r\nfrom positive examples (nodes in the knowledge graph), we perform biased random\r\nwalks and translate them to description logic concepts. Moreover, we improve\r\nsupport for data properties by maximizing information gain when deciding where\r\nto split the data. We show that our approach significantly outperforms the\r\nstate of the art on the benchmarking framework SML-Bench for structured machine\r\nlearning. Our ablation study confirms that this is due to our novel\r\ninitialization method and support for data properties.","lang":"eng"}],"publisher":"ACM","author":[{"id":"11871","last_name":"Heindorf","full_name":"Heindorf, Stefan","first_name":"Stefan","orcid":"0000-0002-4525-6865"},{"full_name":"Blübaum, Lukas","last_name":"Blübaum","first_name":"Lukas"},{"last_name":"Düsterhus","full_name":"Düsterhus, Nick","first_name":"Nick"},{"first_name":"Till","full_name":"Werner, Till","last_name":"Werner"},{"full_name":"Golani, Varun Nandkumar","last_name":"Golani","first_name":"Varun Nandkumar"},{"full_name":"Demir, Caglar","id":"43817","last_name":"Demir","first_name":"Caglar"},{"full_name":"Ngonga Ngomo, Axel-Cyrille","id":"65716","last_name":"Ngonga Ngomo","first_name":"Axel-Cyrille"}],"page":"818-828","type":"conference","date_updated":"2022-10-16T08:49:22Z","publication":"WWW","language":[{"iso":"eng"}],"main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/2111.04879"}]}