Norovirus in the UK Biobank: Silver-Standard Labels, Semi-Supervised Models.

Researching norovirus gastroenteritis retrospectively is challenging, since the absence of a disease record does not imply the absence of that disease. We compared three classification approaches in the UK Biobank: silver-standard labeled controls only, naïve additional unlabeled controls, and semi-supervised models. Naïve approaches classified many unlabeled samples as negative and labeled controls positive. Semi-supervised and labeled models were comparable.