inCognita Major Depressive Disorder (MDD) Results
Suppose we are given access to passively-collected smartphone data for a patient and wish to predict whether the individual has a brain disorder of interest. A supervised learning approach to this task entails:
i.) acquiring smartphone sensor traces for a cohort of patients who have been labeled according to whether they have the target disorder (TD), preferably by clinical experts, and
ii.) using these examples to learn a model which recognizes the TD-status of new (unseen) persons [20]. Unfortunately, in medical passive-sensing applications there are rarely enough labeled examples to support induction of good, generalizable models.