The Predictive Processing Account of Infant Colour Vision Development on Generative Models Revisited
Keywords
No Thumbnail Available
Authors
Issue Date
2019-01-28
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Predictive processing is a theory that states that the human brain is
essentially a prediction machine that tries to predict future sensory input
from the environment. In my thesis I try to partly replicate a previous B.Sc.
thesis experiment that studied developmental predictive processing in which
two di erent scenarios were modelled. The model learns intensities based
on motor movement and colour perception is added afterwards in one sce-
nario. In the other scenario the model immediately learns intensities and
colours based on movements. The changes I made to the experiment are
rstly a more realistic environment and secondly I expanded on the number
of trails I did for the experiment. I adopted the research question of the
experiment I'm replicating: \What is the di erence in the size of the to-
tal prediction error between step-wise learning intensities and colour based
on motor movements (Scenario 1), and immediately learning intensities and
colour perception based on motor movements (Scenario 2) in a more com-
plex environment?". Literature on the development of colour perception in
newborns shows that newborns start by perceiving only intensities and only
during some weeks start discriminating between di erent colours. My expec-
tation is in line with the literature and thus that Scenario 1 will get a lower
total prediction error. I used the Kullback-Leibler divergence to calculate the
total prediction error. Scenario 1 had a total prediction error of 237.3 while
Scenario 2 had a total prediction error of 126.2, so clearly in this experiment
learning intensity and colour directly from motor movements (Scenario 2)
resulted in a lower prediction error.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen