Patrick Fox-Roberts's Biography

Patrick Fox-Roberts, PostDoc, Kings College London

Patrick Fox-Roberts is a research associate at the Randall Divison of King's College London, working on the theoretical and practical limitations of localisation microscopy.

Patrick's background is in machine learning. His PhD at Cambridge University focused on learning statistical models from datasets with missing labels, in particular the curious observation that adding extra unlabelled data to a training set can in some circumstances degrade rather than improve performance. He also showed that in the maximum likelihood case this could be understood by recasting the problem in terms of Kullback-Liebler divergences.

In 2011 he moved to the Cox group at King's College London, and to a focus on applying computer vision and machine learning techniques in biological microscopy. His recent work has centred around exploring better ways of analysing super-resolution localisation microscopy data. In particular, he has been studying the effect of the sample structure itself on the maximum achievable rate at which individual fluorophores can be accurately identified and localised, and investigating the use of machine learning techniques to correct errors that are easily identified by a human but challenging for current algorithms.