Expert deep learning in the human brain: a cognetive computational neuroscience perspective

The human mind has been trained intensively in a wide variety of
complex information processing tasks. In addition, many individuals
develop expertise in specific domains such as chess or microscopy.
Previous studies have supported the hypothesis that expert learners
do not develop super-powers, such as a larger working-memory
capacity, but instead develop a more efficient information processing
that is domain specific (see e.g. Bilalic, 2017). To understand the
underlying representational changes at the neural level, we need to
characterize neural representations in detail, which for a long time
has not been possible due to a lack of suitable experimental and
computational methods. The goal of this project is to investigate the
neural basis of human learning through a state-of-the-art
computational cognitive neuroscience approach. We will compare the
unfolding of representational structure across brain space (functional
magnetic resonance imaging combined with multi-voxel pattern
analyses) and across
time (representational dynamics as studied with multivariate
electroencephalography) between human experts and non-experts,
and between human and artificial expert systems (deep neural
networks). Through this multidisciplinary approach we will obtain an
unprecedented interdisciplinary understanding of how universal
learning and domain-specific expertise alter the human mind and
brain.