{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T21:29:19Z","timestamp":1791408559097,"version":"4.3.3"},"publisher-location":"California","reference-count":0,"publisher":"International Joint Conferences on Artificial Intelligence Organization","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2018,7]]},"abstract":"<jats:p>Artificial Neural Networks are powerful function\n\napproximators capable of modelling solutions to\n\na wide variety of problems, both supervised and\n\nunsupervised. As their size and expressivity increases,\n\nso too does the variance of the model,\n\nyielding a nearly ubiquitous overfitting problem.\n\nAlthough mitigated by a variety of model regularisation\n\nmethods, the common cure is to seek\n\nlarge amounts of training data\u2014which is not necessarily\n\neasily obtained\u2014that sufficiently approximates\n\nthe data distribution of the domain we wish\n\nto test on. In contrast, logic programming methods\n\nsuch as Inductive Logic Programming offer an extremely\n\ndata-efficient process by which models can\n\nbe trained to reason on symbolic domains. However,\n\nthese methods are unable to deal with the variety\n\nof domains neural networks can be applied to:\n\nthey are not robust to noise in or mislabelling of inputs,\n\nand perhaps more importantly, cannot be applied\n\nto non-symbolic domains where the data is\n\nambiguous, such as operating on raw pixels. In this\n\npaper, we propose a Differentiable Inductive Logic\n\nframework (\u2202ILP), which can not only solve tasks\n\nwhich traditional ILP systems are suited for, but\n\nshows a robustness to noise and error in the training\n\ndata which ILP cannot cope with. Furthermore,\n\nas it is trained by backpropagation against a likelihood\n\nobjective, it can be hybridised by connecting\n\nit with neural networks over ambiguous data in\n\norder to be applied to domains which ILP cannot\n\naddress, while providing data efficiency and generalisation\n\nbeyond what neural networks on their own\n\ncan achieve.<\/jats:p>","DOI":"10.24963\/ijcai.2018\/792","type":"proceedings-article","created":{"date-parts":[[2018,7,5]],"date-time":"2018-07-05T01:49:10Z","timestamp":1530755350000},"page":"5598-5602","source":"Crossref","is-referenced-by-count":14,"title":["Learning Explanatory Rules from Noisy Data (Extended Abstract)"],"prefix":"10.24963","author":[{"given":"Richard","family":"Evans","sequence":"first","affiliation":[{"name":"DeepMind"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Edward","family":"Grefenstette","sequence":"additional","affiliation":[{"name":"DeepMind"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"10584","event":{"name":"Twenty-Seventh International Joint Conference on Artificial Intelligence {IJCAI-18}","theme":"Artificial Intelligence","location":"Stockholm, Sweden","acronym":"IJCAI-2018","number":"27","sponsor":["International Joint Conferences on Artificial Intelligence Organization (IJCAI)"],"start":{"date-parts":[[2018,7,13]]},"end":{"date-parts":[[2018,7,19]]}},"container-title":["Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence"],"original-title":[],"deposited":{"date-parts":[[2018,7,5]],"date-time":"2018-07-05T01:56:06Z","timestamp":1530755766000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ijcai.org\/proceedings\/2018\/792"}},"subtitle":[],"proceedings-subject":"Artificial Intelligence Research Articles","short-title":[],"issued":{"date-parts":[[2018,7]]},"references-count":0,"URL":"https:\/\/doi.org\/10.24963\/ijcai.2018\/792","relation":{},"subject":[],"published":{"date-parts":[[2018,7]]}}}