{"links":{"self":"http://dataportal.arc.gov.au/NCGP/API/grants/DE260100189"},"data":{"type":"grant-details","id":"DE260100189","attributes":{"code":"DE260100189","administering-organisation":"Monash University","announcement-administering-organisation":"Monash University","scheme-name":"Discovery Early Career Researcher Award","grant-status":"Active","funding-commencement-year":2026,"years-funded":3,"project-start-date":"2026-03-02","anticipated-end-date":"2029-03-01","grant-summary":"Universal uncertainty quantification using deep learning. This project aims to develop a new and universal approach to uncertainty quantification using deep learning. This project expects to use innovative deep learning tools to develop the first simultaneously tractable and expressive models that can be used directly to quantify uncertainty, a significant unsolved problem. Expected outcomes of this project include a general framework for directly quantifying uncertainty, surpassing current methods which are unable to use big data or are indirect, slow, inexact or inexpressive. This should provide significant benefits for trusted uncertainty quantification using deep learning, with demonstrated downstream applications in manufacturing and coastal bathymetry.","funding-current":524291.00,"funding-at-announcement":520058,"investigators-current":[{"title":"Dr","firstName":"Russell","familyName":"Tsuchida","roleName":"Discovery Early Career Researcher Award","roleCode":"DECRA","isFellowship":true,"orcidIdentifier":null}],"investigators-at-announcement":[{"title":"Dr","firstName":"Russell","familyName":"Tsuchida","roleName":"Discovery Early Career Researcher Award","roleCode":"DECRA","isFellowship":true,"orcidIdentifier":null}],"organisations-current":[{"organisationName":"Monash University","roleName":"Administering Organisation","state":"VIC"}],"organisations-at-announcement":[{"organisationName":"Monash University","roleName":"Administering Organisation","state":"VIC"}],"field-of-research":[{"isPrimary":true,"code":"4611","name":"Machine Learning","type":"FOR20"},{"isPrimary":false,"code":"461103","name":"Deep Learning","type":"FOR20"},{"isPrimary":false,"code":"490508","name":"Statistical Data Science","type":"FOR20"}],"socio-economic-objective":[{"code":"220403","name":"Artificial Intelligence","type":"SEO20"}],"international-collaboration":["England","Japan","United States of America"],"lief-register":[],"achievement-summary":null,"national-interest-test-statement":"Australia's AI and related technologies are valued at potentially AU$315 billion by 2028. AI models need to be able to express uncertainty, confidence or degree of belief about their environment in order to mitigate, communicate and defer to humans potential risks, benefits and possibilities of future actions or predictions. This is especially true when AI models are given partial information collected from multiple sources. However, existing research plans do not adequately address the problem of quantifying uncertainty, and existing methods use outdated or inappropriate tools. This research addresses this gap by building a new AI-based framework that will allow AI models to quantify uncertainty. This framework will be applicable to a range of settings in robotics, manufacturing and natural sciences. Proper treatment of uncertainty will provide economic benefits, by allowing AI models to properly deal with economic risks and benefits. Improved uncertainty quantification will also provide social benefits and improve trust in AI, allowing decision makers to know when they should and should not rely on AI. Research outcomes will be disseminated to the broader community through a workshop, science communication articles, and open-source software. Direct applications in manufacturing and coastal bathymetry will be demonstrated. This research has a high likelihood of being adopted by Australian industry, based on the candidate's previous impact on actors such as Google and CSIRO."}}}