Frederico Parra has defended his thesis “New developments in deep multimodal fusion unleash the automatic assessment of emotion dysregulation in adults”

Abstract
In this work, we develop a novel approach to deep multimodal fusion that
imitates a clinicians’ perception in the assessment of psychological
dimensions, specifically emotion dysregulation in adults. We propose the
multimodal codex as an alternative to early and late multimodal fusion
and show how it allows training deep multimodal fusion models with just a
couple of hundred training examples and only a few thousand parameters.
Our work achievements serve the many fields they touch: to psychometry,
they offer the very first validated multimodal psychometric test for
emotion dysregulation in adults; to multimodal fusion, a new paradigm
that resembles more the way we humans integrate our perceptions; and to
deep learning, a path towards leaner models that limit memorization and
can learn more from less.