Arinbjörn Kolbeinsson is an honorary research associate in the Department of Epidemiology and Biostatistics.
His work focuses on developing machine learning methods that learn from high-dimensional biomedical data including genetic, brain MRI and hospital diagnosis data for personalised health outcome prediction.
Recent research activities include an interpretable deep learning system that learns a biomarker from brain structure that associates with both lifestyle factors, risk factors and diseases. In another a novel prediction model for polygenic scores.
During his PhD, Arinbjörn completed a research internship at Samsung AI and developed tensor methods for improved robustness of deep learning methods in collaboration with Dr Jean Kossaifi and Professor Anima Anandkumar at Nvidia/Caltech.
et al., 2021, Robust deep learning optical autofocus system applied to automated multiwell plate single molecule localization microscopy, Journal of Microscopy, ISSN:0022-2720
et al., 2021, Tensor dropout for robust learning, Ieee Journal of Selected Topics in Signal Processing, Vol:15, ISSN:1932-4553, Pages:630-640
et al., 2020, Accelerated MRI-predicted brain ageing and its associations with cardiometabolic and brain disorders, Scientific Reports, Vol:10, ISSN:2045-2322
et al., 2020, Patch-Based Brain Age Estimation from MR Images, Lecture Notes in Computer Science (including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol:12449 LNCS, ISSN:0302-9743, Pages:98-107