Predicting Depression with Bayesian Nonparametric Models
Keywords
Loading...
Authors
Issue Date
2022-08-01
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
The network model approach is a fairly recent paradigm for modeling psychopathological conditions
like Major Depressive Disorder. This approach models the disorders as an interconnected
network of symptoms. The connections in such a network are usually assumed to be static, however,
there are reasons to disregard this assumption. In this project, we try to apply a Bayesian nonparametric
method called the Generalized Wishart Process (GWP) to model the symptom connections
in the form of a dynamic covariance matrix. We want to investigate whether this method is capable
of learning such dynamic network structures and whether adding the dynamics helps the model to
make better predictions. The GWP was able to learn the dynamic network structures, although it
was worse at it than the parametric baseline with a dynamic covariance. The GWP was also worse
at predicting future symptom values than another Bayesian nonparametric model that used static
covariance. Nevertheless, it showed promising results in terms of interpretability as compared to the
baseline, a property valuable in model prediction.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen