{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/bayesian-heatmaps-probabilistic","title":"Bayesian Heatmaps: Probabilistic Classification with Multiple Unreliable Information Sources","arxiv_id":"1904.03063","date":"2019-04-05","proceeding":null,"authors":["Edwin Simpson","Steven Reece","Stephen J. Roberts"],"abstract":"Unstructured data from diverse sources, such as social media and aerial\nimagery, can provide valuable up-to-date information for intelligent situation\nassessment. Mining these different information sources could bring major\nbenefits to applications such as situation awareness in disaster zones and\nmapping the spread of diseases. Such applications depend on classifying the\nsituation across a region of interest, which can be depicted as a spatial\n\"heatmap\". Annotating unstructured data using crowdsourcing or automated\nclassifiers produces individual classifications at sparse locations that\ntypically contain many errors. We propose a novel Bayesian approach that models\nthe relevance, error rates and bias of each information source, enabling us to\nlearn a spatial Gaussian Process classifier by aggregating data from multiple\nsources with varying reliability and relevance. Our method does not require\ngold-labelled data and can make predictions at any location in an area of\ninterest given only sparse observations. We show empirically that our approach\ncan handle noisy and biased data sources, and that simultaneously inferring\nreliability and transferring information between neighbouring reports leads to\nmore accurate predictions. We demonstrate our method on two real-world problems\nfrom disaster response, showing how our approach reduces the amount of\ncrowdsourced data required and can be used to generate valuable heatmap\nvisualisations from SMS messages and satellite images.","url_abs":"http://arxiv.org/abs/1904.03063v1","url_pdf":"http://arxiv.org/pdf/1904.03063v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"bayesian-heatmaps-probabilistic","repo_url":"https://github.com/OxfordML/heatmap_expts","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"disaster-response","task_name":"Disaster Response"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[{"method_slug":"gaussian-process","method_name":"Gaussian Process"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}