Skip to content Skip to menu
This website uses cookies to help us understand the way visitors use our website. We can't identify you with them and we don't share the data with anyone else. If you click Reject we will set a single cookie to remember your preference. Find out more in UCL's privacy notices.

Multi-view machine learning methods to uncover brain-behaviour associations

The heterogeneity of neurological and mental disorders has been a key confound to disease understanding and treatment outcome prediction, as the patient populations typically include multiple subgroups that do not align with the diagnostic categories. In this talk, I will present a few studies where multivariate methods, such as Canonical Correlation Analysis (CCA), have improved our understanding of the underlying dimensions/factors of disease by uncovering multivariate associations between brain imaging and non-imaging data (e.g., self-report questionnaires and cognitive tests). I will then present probabilistic multi-view methods, such as Group Factor Analysis (GFA), that can be used to address some limitations of CCA, e.g., uncover associations among more than two data modalities or handle missing values. Finally, I will talk about an extension of GFA to identify latent disease factors differently expressed in subgroups of patients.